CN114549074A

CN114549074A - Account recommendation method, device, equipment, storage medium and program product

Info

Publication number: CN114549074A
Application number: CN202210156971.XA
Authority: CN
Inventors: 温灏; 谢乾龙; 王兴星
Original assignee: Beijing Sankuai Online Technology Co Ltd
Current assignee: Beijing Sankuai Online Technology Co Ltd
Priority date: 2022-02-21
Filing date: 2022-02-21
Publication date: 2022-05-27

Abstract

The application discloses an account recommending method, device, equipment, storage medium and program product, and relates to the field of data processing. The method comprises the following steps: acquiring historical flow data of a first identity account; in response to receiving an account number recommendation request sent by a second identity account number, adopting a first recommendation algorithm with a contract service rate as a sequencing core, and sequencing recommendation degrees of the first identity account number through historical flow data to obtain a first identity account number sequence; determining a target recommendation probability of the first identity account based on historical flow data by adopting a second recommendation algorithm taking recommendation effectiveness as a core; a first identity account recommended to a second identity account is determined from the sequence of first identity accounts based on the target recommendation probability. By the aid of the method, computing resources and network resources consumed in an invalid pushing process can be reduced, and invalid account recommendation is avoided. The method and the device can be applied to various scenes such as cloud technology, artificial intelligence, business service and the like.

Description

Account recommendation method, device, equipment, storage medium and program product

Technical Field

The embodiment of the application relates to the field of data processing, in particular to an account recommending method, an account recommending device, account recommending equipment, a storage medium and a program product.

Background

Contract advertisement is a business model based on contracts, the total amount of advertiser demands and the price of orders are specified in the contracts, and the platform is put in a guarantee amount according to the contract specification, so that effective and stable advertisement exposure opportunities are provided for advertisers, and the purpose of reaching users is achieved.

In the related art, the platform typically pushes contract advertisements to users according to the number of pushes specified in the contract signed by the merchant, such as: and (4) the contract advertisements are pushed to more than 5 ten thousand users according to the rules in the contract, and the platform randomly pushes the contract advertisements to the 5 ten thousand users according to the pushing quantity.

However, by the method, only the number of the released contracts is usually considered in the releasing process of the platform for the contracted advertisements, the releasing mode has low accuracy, and the computing resources and the network resources of the platform server are wasted; meanwhile, a large amount of invalid messages which are not interested by the user are easily contained in the contract advertisements received by the user, so that the user frequently refreshes the pushed contract advertisements, and the human-computer interaction efficiency is low.

Disclosure of Invention

The embodiment of the application provides an account recommendation method, an account recommendation device, equipment, a storage medium and a program product, which can reduce computing resources and network resources consumed in an invalid pushing process and avoid invalid account recommendation. The technical scheme is as follows.

In one aspect, an account recommendation method is provided, and the method includes:

acquiring historical flow data of at least one first identity account in a historical time period, wherein the first identity account is used for providing target service;

in response to receiving an account number recommendation request sent by a second identity account number, performing recommendation degree sequencing on the first identity account number through the historical flow data by adopting a first recommendation algorithm to obtain a first identity account number sequence, wherein the first recommendation algorithm is used for sequencing the first identity account number by taking a contract service rate as a sequencing core, and the second identity account number is used for receiving the target service provided by the first identity account number;

determining a target recommendation probability of the at least one first identity account based on the historical traffic data by adopting a second recommendation algorithm, wherein the second recommendation algorithm is used for performing probability prediction on the first identity account by taking recommendation effectiveness as a core;

determining, from the sequence of first identity account numbers, a first identity account number recommended to the second identity account number based on the target recommendation probability.

In another aspect, an account recommending apparatus is provided, the apparatus including:

the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring historical flow data of at least one first identity account in a historical time period, and the first identity account is used for providing target service;

the system comprises a sequencing module, a first recommendation algorithm and a second recommendation algorithm, wherein the sequencing module is used for responding to an account number recommendation request sent by a second identity account number, sequencing recommendation degrees of the first identity account number through historical flow data by adopting the first recommendation algorithm to obtain a first identity account number sequence, the first recommendation algorithm is used for sequencing the first identity account number by taking a contract service rate as a sequencing core, and the second identity account number is used for receiving the target service provided by the first identity account number;

a determining module, configured to determine a target recommendation probability of the at least one first identity account based on the historical traffic data by using a second recommendation algorithm, where the second recommendation algorithm is configured to perform probability prediction on the first identity account with recommendation validity as a core;

and the recommending module is used for determining the first identity account recommended to the second identity account from the first identity account sequence based on the target recommending probability.

In another aspect, a computer device is provided, which includes a processor and a memory, where at least one instruction, at least one program, a set of codes, or a set of instructions is stored in the memory, and the at least one instruction, the at least one program, the set of codes, or the set of instructions is loaded and executed by the processor to implement the account recommendation method according to any one of the embodiments of the present application.

In another aspect, a computer-readable storage medium is provided, in which at least one instruction, at least one program, a set of codes, or a set of instructions is stored, and the at least one instruction, the at least one program, the set of codes, or the set of instructions is loaded and executed by a processor to implement the account number recommendation method according to any one of the embodiments of the present application.

In another aspect, a computer program product or computer program is provided, the computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, so that the computer device executes the account recommendation method in any one of the above embodiments.

The beneficial effects brought by the technical scheme provided by the embodiment of the application at least comprise:

the method comprises the steps of obtaining historical flow data of a plurality of first identity accounts in a historical time period, adopting a first recommendation algorithm with contract service rate as a sequencing core to sequence recommendation degrees of the plurality of first identity accounts, and based on the historical flow data, adopting a second recommendation algorithm with recommendation effectiveness as a core to determine target recommendation probability. The recommendation probability is analyzed under double-layer parallel conditions of contract service rate and recommendation effectiveness, the recommendation accuracy is improved, computing resources and network resources consumed in an invalid pushing process are reduced, the first identity account numbers with enough quantity are recommended to the second identity account numbers, the requirements of the recommended first identity account numbers and the second identity account numbers are matched better, invalid account number recommendation is avoided, and meanwhile the human-computer interaction efficiency of a user is improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a schematic illustration of an implementation environment provided by an exemplary embodiment of the present application;

fig. 2 is a flowchart of an account recommendation method according to an exemplary embodiment of the present application;

FIG. 3 is a flowchart of an account recommendation method according to another exemplary embodiment of the present application;

fig. 4 is a schematic diagram illustrating account recommendation performed by an account recommendation method according to an exemplary embodiment of the present application;

FIG. 5 is a flowchart of an account recommendation method according to another exemplary embodiment of the present application;

fig. 6 is a block diagram illustrating a structure of an account recommending apparatus according to an exemplary embodiment of the present application;

fig. 7 is a block diagram illustrating a structure of an account recommending apparatus according to another exemplary embodiment of the present application;

fig. 8 is a block diagram of a server according to an exemplary embodiment of the present application.

Detailed Description

To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

In the related art, the platform typically pushes contract advertisements to users according to the number of pushes specified in the contract signed by the merchant, such as: and (4) the contract advertisement is pushed to more than 5 thousands of users according to the stipulation in the contract, and the platform randomly pushes the contract advertisement to the 5 thousands of users according to the pushing quantity. However, by the method, only the number of the released contracts is usually considered in the releasing process of the platform for the contracted advertisements, the releasing mode has low accuracy, and the computing resources and the network resources of the platform server are wasted; meanwhile, a large amount of invalid messages which are not interested by the user are easily contained in the contract advertisements received by the user, so that the user frequently refreshes the pushed contract advertisements, and the human-computer interaction efficiency is low.

In the embodiment of the application, the account recommendation method is provided, which can reduce the consumption of computing resources and network resources in an invalid pushing process and avoid invalid account recommendation. The account recommending method obtained through training in the application comprises at least one of the following scenes when in application.

One, apply to under the takeaway service scene

With the development of the take-out industry, more and more users choose to purchase the food, drink and other articles supporting the delivery merchants through the network platform, and the convenience of life is fully improved. The merchant establishes a channel for recommending the merchant to the user by entering the network platform or determining a contract relation with the network platform. When a user selects merchants based on a network platform, the network platform generally provides a plurality of merchants for the user to select, however, some merchants have good public praise but less flow data, or some merchants have unique taste and cannot adapt to all users, and when the above situation occurs, the use experience of the merchants and the users is seriously influenced. Illustratively, by adopting the account number recommendation method, after the recommendation degrees of merchants are ranked based on the historical flow data of the merchants, the target recommendation probability of the first identity account number is determined based on the historical flow data, the merchants are recommended to the user according to the target recommendation probability, so that the situation that only merchants with higher contract service rate are recommended to the user, new merchants with less flow data are ignored, the relationship among different merchants is balanced, the situation that a large number of merchants are pushed inefficiently is avoided, the situation that the user browses too many irrelevant merchants to increase data interaction is also avoided, and the process of recommending more various and more suitable merchants to the user for selection is better realized.

Secondly, the method is applied to random recommendation scenes

When a user opens a content-type application, such as a news application, a video application, or the like, the platform will typically automatically recommend content to the user that the user is interested in or trending for the user to choose to read or view. However, different users do not have the same acceptance of news types or video types, and easily lose interest in the platform when the user receives uninterested content multiple times or even continuously. Illustratively, by taking video recommendation content as an example for explanation, by adopting the account recommendation method, after the recommendation degrees of video publishers are sorted based on historical flow data of the video publishers, the target recommendation probability of a first identity account is determined based on the historical flow data, and the video publishers are recommended to video viewers according to the target recommendation probability, so that the situation that only video publishers with higher contract service rate are recommended to the video viewers can be avoided, newly-resident video publishers with less flow data are ignored, the relationship among different video publishers is balanced, accurate push is performed for the video viewers, on the premise of ensuring the push quantity, the data interaction efficiency is improved, the condition that the demands of the pushed video publishers and the video viewers are not consistent is avoided, and more diverse and more interesting video publishers are provided for selection.

It is to be noted that the application scenario is only an illustrative example, and the account recommendation method provided in this embodiment may also be applied to other scenarios, which are not limited in this embodiment.

It should be noted that information (including but not limited to user equipment information, user personal information, etc.), data (including but not limited to data for analysis, stored data, presented data, etc.), and signals referred to in this application are authorized by the user or sufficiently authorized by various parties, and the collection, use, and processing of the relevant data is required to comply with relevant laws and regulations and standards in relevant countries and regions. For example, traffic data referred to in this application is obtained with sufficient authorization.

Next, an implementation environment related to the embodiment of the present application is described, and referring to fig. 1, schematically, the implementation environment relates to a plurality of first terminals 110, second terminals 120, and a server 130, where the first terminal 110 is connected to the server 130 through a communication network 140, and the server 130 further includes an account recommendation model 150.

In some embodiments, the first terminal 110 is configured to send the historical traffic data to the server 130. Illustratively, the server 130 has a traffic data prediction function, a probability adjustment function, and the like.

The server 130 includes an account recommendation model 150, and after analyzing the historical traffic data by the account recommendation model 150, outputs a recommendation result of the first identity account recommended to the second identity account, and optionally, the server 130 feeds back the recommendation result to the second terminal 120 for display.

Optionally, different first identity accounts log on different first terminals, the plurality of first terminals 110 first send historical traffic data corresponding to the different first identity accounts to the server 130, the plurality of first identity accounts are subjected to recommendation degree sorting through an account recommendation model 150 in the server 130 based on the obtained historical traffic data corresponding to the different first identity accounts to obtain a first account sequence, then, based on the historical traffic data, a second recommendation algorithm is used for determining target recommendation probabilities corresponding to the different first identity accounts, and the first identity accounts are recommended to the second identity accounts with the numerical value of the target recommendation probability as a standard, so that an account recommendation process is realized. Optionally, the server 130 feeds back the recommendation result to the second terminal 120 for display.

The above process is an example of a non-exclusive case where the account recommendation model 150 applies the process.

It should be noted that the above terminals include, but are not limited to, mobile terminals such as mobile phones, tablet computers, portable laptop computers, intelligent voice interaction devices, intelligent home appliances, and vehicle-mounted terminals, and can also be implemented as desktop computers; the server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as cloud service, a cloud database, cloud computing, a cloud function, cloud storage, Network service, cloud communication, middleware service, domain name service, security service, Content Delivery Network (CDN), big data and an artificial intelligence platform.

In some embodiments, the servers described above may also be implemented as nodes in a blockchain system.

With reference to the above noun introduction and application scenarios, the account recommendation method provided in the present application is described, taking the application of the method to a server as an example, as shown in fig. 2, the method includes the following steps 210 to 240.

Step 210, obtaining historical flow data of at least one first identity account in a historical time period.

Wherein the first identity account is used for providing a target service.

Illustratively, the target service includes multiple types such as a dining service, a beauty service, an entertainment service, and the like, and according to the type difference of the target service, the first identity account is correspondingly presented in different forms, for example: providing a first identity account of the catering service as a catering merchant; providing a first identity account of beauty service as a beauty merchant; the first identity account providing entertainment services is an entertainment merchant, etc.

Optionally, the first identity account is represented by a name, such as: the first identity account is the account of store A; alternatively, the first identity account number is represented by a featured product, such as: the first identity account is represented by a characteristic B milk tea and the like.

Illustratively, the historical time period is used to indicate a certain time period before the current time, such as: presetting a historical time period as one week before the current time; alternatively, the historical time period is predetermined to be the day before the current time, and the like.

In an alternative embodiment, the historical traffic data includes at least one of click-through rate data, conversion rate data, and customer price data.

Wherein the click-through rate data is related to a number of times the first identity account was clicked. Illustratively, the click rate data is the number of times the first identity account is clicked; alternatively, the click rate data is the quotient of the number of clicks and the number of exposures of the first identity account.

Optionally, the manner in which the first identity account number is clicked includes at least one of the following.

1. And clicking a corresponding area of the first identity account.

Schematically, the first identity account is taken as the C account for explanation. The account C is a food account, the related information of the account C is disclosed on the platform D by the account C, so that users using the platform D and the platform D can know the related information of the account C, and the account C has a corresponding area on the platform D. For example: and when the user of the D platform selects the food, displaying a food recommendation column on the terminal of the user, wherein the food recommendation column comprises a plurality of first identity accounts. Each first identity account number has a corresponding area, and the corresponding area is used for displaying related information such as names, positions, characteristic products and the like corresponding to different first identity account numbers. Illustratively, the C account has a corresponding C area, and after a user of the D platform clicks the C area, the process of clicking the C account is realized.

2. And triggering a sharing control of the first identity account.

Illustratively, different first identity accounts have corresponding home page spaces, and the home page spaces include more detailed related information such as the name, the location, the evaluation information, the feature product, and the like of the first identity account. Optionally, the home page space includes a sharing control for sharing to other friends in a sharing form such as a link or a text, for example: the user shares the sharing link with friends (such as friends in a friend list) related to the user on the platform logged in by the first identity account, or shares the sharing link with other social platform friends related to the D platform, and the like. Optionally, the C account logs on a D platform, a C homepage space corresponding to the C account is arranged in the D platform, a user of the D platform triggers a sharing control in the C homepage space to generate a sharing link corresponding to the C account, the user can share the sharing link to other friends, and the other friends can click the sharing link to realize a click process of the C account; or after the user of the D platform triggers the sharing control in the C homepage space, the text sharing content corresponding to the C account is generated, the user can paste the text sharing content into a chat box and send the content to other friends, and the other friends copy the text sharing content and open the D platform to realize the click process of the C account.

It should be noted that the above is only an illustrative example, and the present invention is not limited to this.

Optionally, the conversion rate data in the traffic data is related to the number of clicks and the number of next orders of the first account number, and illustratively, the conversion rate data is the quotient of the number of next orders and the number of clicks of the first account number.

Wherein the order number is used for indicating the times of purchasing, reserving and the like of the product in the first identity account number. Illustratively, the first identity account is a C account, the C account is a food merchant and includes various foods, the C account logs on the D platform, and a user of the D platform performs purchase operation, reservation operation, and the like on at least one of the various foods under the C account, so as to implement an ordering process for the C account.

Optionally, the customer price data in the traffic data is used to indicate a price each user consumes for the first identity account. Illustratively, the account C is a food merchant, which includes a plurality of foods, and the customer price data is the quotient of sales and user number (customer number). For example: the historical time period is 1 day, the sales amount of the acquired C account in the past day is 2450 yuan, the number of customers is 20, and the customer price data of the C account in the past day is 122.5 yuan.

In an alternative embodiment, the traffic data may be implemented as sharing rate data (e.g., the number of times of being shared, or the quotient of the number of times of being shared and the number of times of being clicked, etc.), transaction rate data (e.g., the number of successful transactions, or the quotient of the number of successful transactions and the number of times of placing orders, etc.), or the like, in addition to the click rate data, the conversion rate data, and the customer price data. The above description is merely exemplary, and the present disclosure is not limited thereto.

Step 220, in response to receiving an account number recommendation request sent by the second identity account number, performing recommendation degree sequencing on the first identity account number through historical flow data by using a first recommendation algorithm to obtain a first identity account number sequence.

The second identity account is used for receiving the target service provided by the first identity account, and the account recommendation request is used for indicating the second identity account to acquire the request of the first identity account. Illustratively, the second identity account and the first identity account are accounts logged in the same platform, the first identity account is an account logged in the platform and providing the target service, and the second identity account is an account logged in the platform and seeking the target service. For example: the platform provides online service for users through an M application program, a C merchant signs a contract with the platform to realize a process of entering the platform, and the C merchant presents a C account in the M application program; the D user obtains a D account number by registering on the M application program, and the D account number is presented in the forms of a default ID, a registered number (such as a mobile phone number), a nickname and the like.

Optionally, the sending method for sending the account recommendation request by the second identity account includes at least one of the following methods.

(1) The process of sending the account number recommendation request is realized by entering a platform where the first identity account number is resident.

Illustratively, the first identity account is resident on the M platform or has a contract with the M platform, and the second identity account realizes a process of sending an account recommendation request when opening an M application program corresponding to the M platform.

(2) And triggering the specified area to realize the process of sending the account recommendation request.

Illustratively, the first identity account is resident in an M platform, and the second identity account triggers a designated area in an M application program after the M application program corresponding to the M platform is opened, so as to implement a process of sending an account recommendation request. For example: in the M application program, a plurality of different areas for realizing different functions are correspondingly arranged, such as a code scanning area, a card ticket area, a food area, a performance area and the like, and the code scanning function is started in response to the triggering operation of the code scanning area; or responding to the triggering operation of the 'food' area, realizing the process of sending the account number recommendation request and the like.

It should be noted that the above are only exemplary, and the embodiments of the present application are not limited thereto.

Optionally, the first recommendation algorithm is configured to rank the first identity account with a contract service rate as a ranking core, where the contract service rate is used to indicate a case where the first identity account is recommended to the second identity account.

Illustratively, the contract is a defined contract between the first identity account and the platform, and is used for instructing the platform to recommend the first identity account to the second identity account. For example: and signing a contract between the first identity account and the platform, wherein the contract specifies that the platform recommends the first identity account to the second identity account by the platform after the platform is opened by the second identity account.

Illustratively, the service rate includes the following form: (1) the probability of providing a service, which means the probability of providing a service from a first identity account to a second identity account, is as follows: the probability of recommending the first identity account to the second identity account is 0.8; (2) the number of clicks represents the quotient of the number of clicks and the number of recommendations of the first identity account by the plurality of second identity accounts after the first identity account is recommended to the plurality of second identity accounts; (3) the conversion probability indicates the quotient of the next number of clicks and the like of the second identity account to the first identity account after the first identity account is recommended to the second identity account.

Illustratively, the service rate is taken as an example of the probability of providing the service, and accordingly, the contract service rate is used to indicate the probability of recommending the first identity account to the second identity account. Optionally, the contract service rate is determined in at least one of the following manners.

(1) The contract service rate is predetermined according to the contract terms.

Illustratively, a contract service rate is predefined in a contract, and the contract service rate corresponding to the first identity account number is different according to the difference that the first identity account number and the platform sign the contract. For example: in a contract signed by a first identity account M and a platform, a specified contract service rate is M, in a contract signed by a first identity account N and the platform, a specified contract service rate is N, wherein M is less than N, when the platform recommends the first identity account to a second identity account according to the contract service rate, the probability that the first identity account N with a higher contract service rate is recommended to the second identity account is greater than the probability that the first identity account M with a lower contract service rate is recommended to the second identity account, namely: the platform tends to recommend the first identity account N to the second identity account, as opposed to recommending the first identity account M to the second identity account. Alternatively, different contracts may correspond to the same contract service rate.

(2) And determining a contract service rate according to the historical flow data of the first identity account.

Illustratively, a plurality of first identity account numbers and the platform sign contracts, and the contracts specify the platform event, and contract service rates corresponding to different first identity account numbers are not predetermined. When a first identity account recommended to a second identity account is determined from a plurality of first identity accounts, historical flow data of each first identity account is analyzed on the basis of comprehensively considering the historical flow data corresponding to the plurality of first identity accounts, and then a contract service rate corresponding to each first identity account is determined.

For example: after the historical flow data corresponding to the plurality of first identity account numbers are subjected to mean value operation, average data are obtained, the deviation between the historical flow data of each first identity account number and the average data is analyzed, and then the contract service rate corresponding to each first identity account number is determined, for example: determining a reference contract service rate to be 0.5, determining contract service rates corresponding to different first identity account numbers by taking the reference contract service rate as a deviation center, and when the historical flow data of the first identity account number M is lower than average data and has smaller deviation, determining the contract service rate of the first identity account number M to be a numerical value which is smaller than 0.5 but has a smaller difference with 0.5; when the historical flow data of the first identity account N is higher than the average data and has a larger deviation, the contract service rate of the first identity account N is a numerical value which is larger than 0.5 but has a larger difference from 0.5, and the like.

In an optional embodiment, based on the historical flow data, a first recommendation algorithm is adopted to adjust the estimated flow data corresponding to each first identity account in at least one first identity account, and an adjustment result corresponding to each first identity account is determined.

The estimated flow data is flow data obtained by predicting first identity accounts in a future time period, and each first identity account has corresponding estimated flow data, namely: for different first identity account numbers, estimated flow data such as different click rates, conversion rates, customer unit prices and the like can be estimated. Optionally, before recommending the first identity account to the second identity account, a contract service rate for recommending the first identity account to the second identity account is predetermined.

Illustratively, different first identity account numbers correspond to different contract service rates, the contract service rates include both preset parameters and parameters determined based on historical flow data of the first identity account numbers, and the probability of recommending the first identity account number to the second identity account number is determined through the contract service rates.

In an optional embodiment, based on historical flow data, determining an adjustment parameter for adjusting the predicted flow data; and adjusting the pre-estimated flow data corresponding to each first identity account according to the adjustment parameters and the contract service rate corresponding to each first identity account in at least one first identity account, and determining the adjustment result corresponding to each first identity account.

Illustratively, when the estimated flow data corresponding to each first identity account is adjusted based on the contract service rate, the estimated flow data is adjusted to different degrees according to the difference of the contract service rates corresponding to different first identity accounts.

Alternatively, an exemplary first recommendation algorithm is as follows.

y_j＝α_j(a*pCTR_j+b*pCVR_j+*pPrice)

Wherein, y_jA ranking score for indicating a first identity account number j; a. b and c are adjustment parameters of a first recommendation algorithm related to historical flow data of the first identity account; pCTR_jA pre-estimated click rate value used for indicating the first identity account number j; pCVR_jA conversion rate pre-evaluation value used for indicating the first identity account number i; pPrice_jThe customer order pre-evaluation value is used for indicating the first identity account number j; alpha is alpha_jIndicating a contract service rate for the first identity account number j. The click rate pre-estimated value, the conversion rate pre-estimated value and the passenger unit price pre-estimated value are collectively referred to as pre-estimated flow data.

Optionally, a is an adjustment parameter of the estimated click rate value corresponding to the first identity account, and is related to the historical click rate value in the historical traffic data, and is used to adjust the estimated click rate value, for example: a is the average of a plurality of first identity account click rate history values.

Optionally, b is an adjustment parameter of the conversion rate predicted value corresponding to the first identity account, and is related to the click rate history value and the conversion rate history value in the historical traffic data, and is used to adjust the conversion rate predicted value, for example: b is a weighted average of the plurality of first identity account click rate historical values and the plurality of first identity account conversion rate historical values, namely: and determining an adjusting parameter b and the like according to the influence of the click rate history value of each first identity account in all the first identity accounts and the influence of the conversion rate history value of each first identity account in all the first identity accounts.

In an alternative embodiment, the above-mentioned exemplary first recommendation algorithm is used to determine the ranking score of each first identity account, and then, according to the ranking score of each first identity account, a sequence of the first identity accounts is determined, such as: and according to the numerical value of the sorting score, performing descending order on the sorting score to obtain a first identity account sequence and the like.

In an alternative embodiment, the sequence of first identity accounts is determined based on the adjustment result corresponding to each first identity account.

Illustratively, the first recommendation algorithm is a sorting formula, and the following sorting formula is adopted to sort different first identity account numbers.

RankScore_j＝f(pCTR_j，pCVR_j，pPrice_j，α_j，…)

Wherein, RankScore_jThe system is used for indicating the sorting condition of the first identity account number j; f is used to indicate a ranking function; pCTR_jA pre-estimated click rate value used for indicating the first identity account number j; pCVR_jA conversion estimate for indicating the first identity account number j; pPrice_jA customer order estimate for indicating a first identity account number j; alpha is alpha_jIndicating a contract service rate for the first identity account number j. The click rate pre-estimated value, the conversion rate pre-estimated value and the passenger unit price pre-estimated value are collectively referred to as pre-estimated flow data.

Illustratively, "is used to indicate other estimated flow data, namely: the parameters in the above sorting formula are not limited to the estimated traffic data corresponding to the click rate estimated value, the conversion rate estimated value and the passenger unit price estimated value, and other estimated traffic data may also be used to sort different first identity accounts, for example: sorting different first identity accounts by adopting a sharing rate estimation value, wherein the sharing rate is related to the number of times that the first identity accounts are shared; or, sorting different first identity accounts by adopting a plurality of estimated flow data such as a sharing rate estimated value, a passenger order estimated value and the like.

Illustratively, by using the first recommendation algorithm, a sorting process for sorting different first identity account numbers is implemented based on historical traffic data and contract service rates of the different first identity account numbers.

And step 230, determining a target recommendation probability of the at least one first identity account based on the historical flow data by adopting a second recommendation algorithm.

And the second recommendation algorithm is used for performing probability prediction on the first identity account by taking recommendation effectiveness as a core. The recommendation validity is used for indicating whether the second identity account makes a corresponding response to the first identity account after the first identity account is recommended to the second identity account. For example: after the first identity account is recommended to the second identity account, the second identity account carries out click operation on the first identity account, so that the click operation is regarded as effective recommendation and meets the requirement of recommendation effectiveness; or after the first identity account is recommended to the second identity account, the second identity account performs order placing operation on the first identity account, so that the order placing operation is regarded as effective recommendation, and the recommendation effectiveness requirement is met.

Optionally, an average value operation is performed on historical flow data of at least one first identity account to obtain historical average data.

Illustratively, after obtaining the historical flow data of at least one first identity account, the historical flow data of each first identity account is analyzed respectively. Optionally, when the C account is analyzed, different types of historical traffic data in the historical traffic data corresponding to the C account are analyzed respectively. For example: determining historical flow data of the account C in a historical time period, and performing mean operation on click rate data in the historical flow data to obtain an average value of the click rate data; carrying out mean operation on the conversion rate data in the historical flow data to obtain the average value of the conversion rate data; and carrying out mean operation on the passenger unit price data in the historical traffic data to obtain the mean value of the passenger unit price data. The average value of the click rate data, the average value of the conversion rate data, and the average value of the customer price data are collectively referred to as history average data.

In an optional embodiment, the second recommendation algorithm includes a preset adjustment coefficient, and the historical average data and the estimated traffic data of the ith first identity account in the future time period are analyzed based on the preset adjustment coefficient to obtain the target recommendation probability of the ith first identity account.

The estimated flow data is flow data obtained by predicting the ith first identity account in a future time period, wherein i is a positive integer.

Optionally, the preset adjustment coefficient includes a first preset coefficient and a second preset coefficient.

Illustratively, the first preset coefficient and the second preset coefficient are coefficients determined through a plurality of tests based on historical flow data and predicted flow data. For example: and taking the A moment as a demarcation point, obtaining historical flow data of a certain historical time period before the A moment, and estimating the flow data of a certain future time period after the A moment by using the historical flow data of the historical time period to obtain estimated flow data. And determining a loss value based on the real flow data of the future time period after the time A and the estimated flow data aiming at the future time period, and determining a first preset coefficient and a second preset coefficient for adjusting the contract service rate by taking the loss value as a target.

In an alternative embodiment, the product of the first predetermined coefficient and the historical average data is determined to obtain the first data.

Optionally, the second recommended algorithm is a linear transformation formula, that is: the target recommendation probability is changed along with the preset coefficient. Illustratively, after the historical average data is processed by a first preset coefficient, first data is obtained.

In an optional embodiment, a product of a second preset coefficient and the estimated flow data corresponding to the ith first identity account is determined to obtain second data, where i is a positive integer.

Illustratively, different first identity account numbers are respectively analyzed by adopting a second recommendation algorithm. For example: when the C account is analyzed, after the estimated flow data corresponding to the C account is determined based on the historical flow data corresponding to the C account, the second preset coefficient is multiplied by the estimated flow data corresponding to the C account to obtain second data. Optionally, when the second preset coefficient is multiplied by the estimated flow data corresponding to the account C, and the estimated flow data of multiple types corresponding to the second preset coefficient and the account C are multiplied by each other. For example: and multiplying the second preset coefficient by the click rate predicted value corresponding to the C account, multiplying the second preset coefficient by the conversion rate predicted value corresponding to the C account, and the like. Illustratively, data corresponding to multiple types of estimated flow data obtained through the multiplication operation are collectively referred to as second data.

In an optional embodiment, the contract service rate of the ith first identity account is adjusted by adding the first data and the second data, and the target recommendation probability of the ith first identity account is determined.

The contract service rate is used to indicate a probability of recommending the first identity account to the second identity account. Illustratively, when the contract service rate of the ith first identity account is adjusted by a numerical value obtained by adding the first data and the second data, the target recommendation probability is obtained by multiplying the numerical value obtained by adding the first data and the second data by the contract service rate of the first identity account; or, the target recommendation probability is determined based on a weight ratio of the historical flow data affecting the first data and the predicted flow data affecting the second data, for example: the first identity account is a new store, the historical weight of first data corresponding to historical flow data of the new store is set to be small, the estimated weight of second data corresponding to estimated flow data of the new store is set to be large, the product of the first data and the historical weight is multiplied by the product of the second data and the estimated weight, and then the product is added to obtain target recommendation probability and the like.

Optionally, after obtaining the historical traffic data and the estimated traffic data, adjusting the contract service rate by using a preset adjustment coefficient, where an adjustment formula for adjusting the contract service rate is schematically shown as follows:

wherein,

the method comprises the steps of indicating an adjustment condition after adjusting the preset matching probability of a first identity account j; g is used to indicate an adjustment function; pCTR_jA pre-estimated click rate value used for indicating the first identity account number j; pCVR_jFor indicating a conversion estimate for the first identification number j; pPrice_jA customer order estimate for indicating a first identity account number j; alpha is alpha_jFor indicating a contract service rate for the first identity account number j; CTR (China railway radio)_avgA history value indicating a click rate of the first identity account number j; pCVR_jA historical value indicating a conversion rate of the first identification number j; pPrice_jIndicating a customer order history value for the first identity account number j. The historical click rate values, the historical conversion rate values and the historical passenger order values are collectively referred to as historical traffic data.

Illustratively, the higher the value of the target recommendation probability is, the higher the probability of recommending the first identity account to the second identity account is; the smaller the value of the target recommendation probability, the smaller the probability of recommending the first identity account to the second identity account. Optionally, the relationship between different first identity account numbers can be better coordinated through the target recommendation probability, and the first identity account number can be more flexibly recommended to the second identity account. For example: when the contract service rate is adjusted, the target recommendation probability of the first identity account with a larger value in the control contract service rate does not infinitely tend to 1, and the target recommendation probability of the first identity account with a smaller value in the control contract service rate does not infinitely tend to 0, namely: the relationship between different first identity accounts is balanced as much as possible, not only because the historical traffic data is better, but also the better newly-enrolled first identity account is ignored, and the like.

In step 240, a first identity account recommended to a second identity account is determined from the sequence of first identity accounts based on the target recommendation probability.

The target recommendation probability is used for indicating the probability of recommending the first identity account to the second identity account, which is obtained after the contract service rate is adjusted.

In an optional embodiment, the first k first identity account numbers are intercepted from a first identity account number sequence to obtain a sequence segment, wherein k is a positive integer; and determining a first identity account recommended to the second identity account from the sequence segments based on the target recommendation probability.

Illustratively, after obtaining the first sequence of identity accounts, intercepting the first sequence of identity accounts to obtain a plurality of first identity accounts belonging to the first sequence of identity accounts, for example: the method comprises the steps of obtaining k first identity account numbers in advance, taking the first identity account number in a first identity account number sequence as an interception starting point, and continuously obtaining the k first identity account numbers, so that a sequence segment containing the k first identity account numbers is obtained.

In an optional embodiment, the target recommendation probability of the first identity account is matched with the probability condition to obtain a matching result, wherein the matching result comprises a matching success result; and determining the first identity account number which accords with the matching success result from the sequence fragment as the first identity account number recommended to the second identity account number.

The probability condition comprises a preset condition and a condition determined according to the sequencing condition of the first identity account sequence. Optionally, the probability condition includes at least one of a probability numerical condition and a probability threshold condition, and accordingly, when the probability condition is the probability numerical condition, the matching success result is used to indicate a result with a larger value in the probability numerical values; and when the probability condition is a probability threshold condition, the matching success result is used for indicating the result that the probability reaches the probability threshold condition.

Illustratively, after the first identity account sequence and the target recommendation probability are determined, the first identity account is recommended to the second identity account according to the target recommendation probability. For example: recommending a first identity account with higher target recommendation probability to a second identity account according to the numerical value of the target recommendation probability on the basis of presenting a plurality of first identity accounts in a first identity account sequence; or recommending the first identity account with the target recommendation probability exceeding a preset probability threshold to the second identity account, and the like. The above description is only exemplary, and the present invention is not limited to the above description.

In summary, historical traffic data of the plurality of first identity accounts in a historical time period is obtained, a first recommendation algorithm with contract service rate as a ranking core is adopted to rank recommendation degrees of the plurality of first identity accounts, and a second recommendation algorithm with recommendation effectiveness as a core is adopted to determine a target recommendation probability based on the historical traffic data. The recommendation probability is analyzed under double-layer parallel conditions of contract service rate and recommendation effectiveness, the recommendation accuracy is improved, computing resources and network resources consumed in an invalid pushing process are reduced, the first identity account numbers with enough quantity are recommended to the second identity account numbers, the requirements of the recommended first identity account numbers and the second identity account numbers are matched better, invalid account number recommendation is avoided, and meanwhile the human-computer interaction efficiency of a user is improved.

In an optional embodiment, the first recommendation algorithm and the second recommendation algorithm perform an account recommendation process through estimated traffic data obtained by estimating traffic data. Illustratively, the embodiment shown in steps 210 to 240 can also be implemented as the following steps 310 to 350.

Step 310, obtaining historical flow data of at least one first identity account in a historical time period.

Wherein the first identity account is used for providing a target service.

Illustratively, the target services include various types of food service, beauty service, entertainment service, and the like, such as: providing a first identity account of the catering service as a catering merchant; providing a first identity account of beauty service as a beauty merchant; the first identity account providing entertainment services is an entertainment merchant, etc.

Optionally, the historical traffic data includes at least one of click through rate data, conversion rate data, and customer price data.

Optionally, the conversion rate data is related to the number of clicks and the number of orders of the first account number, and illustratively, the conversion rate data is the number of orders of the first account number and the number of clicks. The customer price data is used to indicate a price to be consumed for each user for the first identity account.

In an optional embodiment, the traffic data may be implemented as sharing rate data, deal rate data, and the like, in addition to the click rate data, the conversion rate data, and the customer price data.

And 320, predicting estimated flow data of at least one first identity account in a future time period based on the historical flow data.

Illustratively, the historical traffic data includes click rate data, conversion rate data, and customer price data of the first identity account over a historical period of time.

In an alternative embodiment, first historical traffic data for a first historical time period and second historical traffic data for a second historical time period of the historical traffic data are obtained.

Optionally, the historical traffic data includes a plurality of traffic data in a plurality of historical time periods, wherein the first historical traffic data and the second historical traffic data are included in the historical traffic data. The relationship between the first historical period of time and the second historical period of time includes at least one of the following.

1. The first historical time period and the second historical time period have a time period coincidence relation

Illustratively, the first historical time period includes a second historical time period, wherein the first historical time period is a historical time period with a larger time range, and the second historical time period is a part of the historical time period in the first historical time period. For example: randomly selecting a time shaft, wherein the time in the time shaft is sequentially represented as point A time, point B time and point C time, the first historical time period is the historical time period from the point A time to the point C time, and the second historical time period is the historical time period from the point A time to the point B time; or the second history time period is the history time period from the time point B to the time point C.

2. The first historical time period and the second historical time period do not have a time period coincidence relation

Illustratively, the first history period and the second history period do not coincide on the time axis. For example: and a time shaft is arbitrarily selected, wherein the time in the time shaft is sequentially represented as A-point time, B-point time, C-point time and D-point time, the first historical time period is a historical time period from the A-point time to the B-point time, and the second historical time period is a historical time period from the C-point time to the D-point time.

In an optional embodiment, the candidate predictive model is used to predict the flow data of the first historical time period, so as to obtain predicted flow data corresponding to the first historical flow data.

And the candidate estimation model is a data model to be trained. Illustratively, the candidate prediction model is a general model with a certain data prediction function, and a data analysis result obtained based on M data prediction can be obtained by inputting M data into the candidate prediction model. For example: and (4) inputting the passenger flow data of N intersections in the next month into the candidate estimation model to obtain a prediction result of the passenger flow data in the next day or week.

In an optional embodiment, the loss value is determined based on first historical flow data and predicted flow data for the at least one first identity account over a first historical period of time; training the candidate estimation model by using the loss value; and responding to the training of the candidate prediction model to achieve a training effect, and obtaining a data prediction model.

Illustratively, training the candidate prediction model by using the loss value obtains the data prediction model because the training of the candidate prediction model reaches the training target, and illustratively, the training target at least includes at least one of the following situations.

1. And responding to the fact that the loss value reaches a convergence state, and taking a candidate estimation model obtained by the last iteration training as a data estimation model.

Illustratively, the reaching of the convergence state by the loss value is used to indicate that the value of the loss value obtained by the loss function is no longer changing or the change amplitude is smaller than a preset threshold value. For example: and the loss value corresponding to the nth group of flow data is 0.1, the loss value corresponding to the (n + 1) th group of flow data is 0.1, the loss value can be considered to reach a convergence state, and the candidate prediction model adjusted by the loss value corresponding to the nth group of flow data or the (n + 1) th group of flow data is used as a data prediction model to realize the training process of the candidate prediction model.

2. And in response to the fact that the obtaining times of the loss value reach a time threshold value, taking a candidate estimation model obtained by the last iteration training as a data estimation model.

Schematically, a loss value can be obtained by one-time obtaining, the obtaining times of the loss value for training the candidate prediction model are preset, and when a group of flow data corresponds to a loss value, the obtaining times of the loss value is the group number of the flow data; or, when a group of flow data corresponds to a plurality of loss values, the number of times of obtaining the loss is the number of the loss values. For example: a loss value can be obtained by presetting one-time obtaining, wherein the time threshold of the loss value obtaining is 10 times, namely when the obtaining time threshold is reached, a candidate prediction model adjusted by the loss value of the last time is used as a data prediction model, or a candidate prediction model adjusted by the minimum loss value in the loss value 10-time adjusting process is used as a data prediction model, and the training process of the candidate prediction model is realized.

In an optional embodiment, the candidate prediction model is trained based on the first historical flow data and the predicted flow data to obtain a data prediction model.

The data prediction model is used for predicting flow data of at least one first identity account in a future time period. Schematically, the click rate data of the first identity account in a future time period is predicted through a data prediction model, so that prompt information for improving the click rate is provided for the first identity account; or through a data prediction model, predicting conversion rate data of the first identity account in a future time period, thereby providing prompt information for improving the conversion rate for the first identity account; or the passenger unit price data of the first identity account in the future time period is predicted through the data prediction model, so that prompt information for reducing or improving the passenger unit price is provided for the first identity account, and the like.

In an optional embodiment, second historical flow data corresponding to at least one first identity account is input into the data prediction model to obtain predicted flow data.

Optionally, after the data estimation model is obtained, the historical flow data corresponding to the first identity account is analyzed based on the data estimation model. Illustratively, the account C to be analyzed is an account of a beauty shop, historical flow data corresponding to the account of the beauty shop are click rate data, conversion rate data and guest unit price data of the beauty shop within the past month, and the historical flow data within the month corresponding to the beauty shop is input into a data estimation model based on a data estimation model obtained through pre-training to obtain estimated flow data.

Optionally, the data prediction model is a prediction model capable of setting a prediction time. Illustratively, the estimated time of the data estimation model is set as one day, the data estimation model can estimate the traffic data of the future one day based on the historical traffic data of the first identity account, and the estimated data is estimated traffic data; or setting the estimated time of the data estimation model as one week, the data estimation model can estimate the flow data of one week in the future based on the historical flow data of the first identity account, and the estimated data is the estimated flow data and the like.

Optionally, the data prediction model is a prediction model capable of performing real-time prediction. Illustratively, historical flow data are updated in real time along with the time, the historical flow data obtained through real-time updating are input into a data prediction model to obtain predicted flow data obtained through real-time prediction, and the process of continuously predicting the flow data in the future time period through the updated historical flow data is achieved.

Optionally, the data prediction model is a prediction model that can be set in a prediction flow data form. The flow data form includes the above-mentioned flow data such as click rate, conversion rate, customer unit price, etc. Illustratively, the flow data form of the data estimation model is set to be respectively estimated, and then the data estimation model can respectively estimate different forms of historical flow data based on the historical flow data of the first identity account. For example: inputting the flow data form of the data estimation model into click rate flow data and conversion rate flow data of the C account in a historical time period, and based on an estimation method for respectively estimating the historical flow data in different forms, outputting the data estimation model into click rate flow data and conversion rate flow data in a future time period estimated by the C account, wherein the estimated click rate flow data and conversion rate flow data are collectively referred to as estimated flow data; or setting the flow data form of the data estimation model as comprehensive estimation, and the data estimation model can comprehensively estimate different forms of historical flow data based on the historical flow data of the first identity account. For example: the flow data input into the data prediction model is in the form of click rate flow data and guest unit price flow data of the C account in a historical time period, the data prediction model outputs a comprehensive score in a future time period predicted by the C account based on the prediction method for comprehensively predicting the historical flow data in different forms, the comprehensive score is obtained based on the click rate flow data and the conversion rate flow data (if the comprehensive score exceeds 7 minutes, the running condition of the C account in the future time period is proved to be better), and the predicted comprehensive score is called as predicted flow data.

Step 330, in response to receiving an account number recommendation request sent by the second identity account number, performing recommendation degree sequencing on the first identity account number by using a first recommendation algorithm through historical flow data and estimated flow data to obtain a first identity account number sequence.

The second identity account is used for receiving the target service provided by the first identity account, and the account recommendation request is used for indicating the second identity account to acquire the request of the first identity account. Illustratively, the second identity account and the first identity account are accounts logged in the same platform, the first identity account is an account logged in the platform and providing the target service, and the second identity account is an account logged in the platform and seeking the target service.

The first recommendation algorithm is used for sequencing the first identity account by taking the contract service rate as a sequencing core. Optionally, based on the historical flow data, adjusting the estimated flow data corresponding to each first identity account in the at least one first identity account by using a first recommendation algorithm, and determining an adjustment result corresponding to each first identity account; after the estimated flow data corresponding to each first identity account is adjusted according to the contract service rate, the first identity account sequence is determined based on the adjustment result corresponding to each first identity account.

Step 340, determining a target recommendation probability of the at least one first identity account based on the historical traffic data and the estimated traffic data by using a second recommendation algorithm.

And the second recommendation algorithm is used for performing probability prediction on the first identity account by taking recommendation effectiveness as a core. Optionally, the second recommendation algorithm includes a preset adjustment coefficient, and the preset adjustment coefficient includes a first preset coefficient and a second preset coefficient obtained based on the historical flow data and the estimated flow data.

Illustratively, performing mean value operation on historical flow data of at least one first identity account to obtain historical mean data; determining a product of a first preset coefficient and historical average data to obtain first data; and determining the product of a second preset coefficient and the estimated flow data corresponding to the ith first identity account to obtain second data, wherein i is a positive integer.

In an optional embodiment, the contract service rate of the ith first identity account is adjusted by a value obtained by adding the first data and the second data, and the target recommendation probability of the ith first identity account is determined.

Illustratively, a value obtained by adding the first data and the second data is multiplied by a contract service rate of the first identity account to obtain a target recommendation probability; or determining the target recommendation probability based on the weight proportion of the historical flow data influencing the first data and the predicted flow data influencing the second data.

Step 350, determining the first identity account number recommended to the second identity account number from the first identity account number sequence based on the target recommendation probability.

Illustratively, the higher the value of the target recommendation probability is, the higher the probability of recommending the first identity account to the second identity account is; the smaller the value of the target recommendation probability, the smaller the probability of recommending the first identity account to the second identity account. The target recommendation probability is, for example: when the contract service rate is adjusted, the target recommendation probability of the first identity account with a larger value in the control contract service rate does not infinitely tend to 1, and the target recommendation probability of the first identity account with a smaller value in the control contract service rate does not infinitely tend to 0, namely: the relationship between different first identity accounts is balanced as much as possible, not only because historical traffic data is better, but also better new resident first identity accounts and the like are ignored.

In an optional embodiment, determining a target recommendation probability corresponding to each first identity account; and matching the numerical value of the target recommendation probability with the probability condition, and determining the first identity account recommended to the second identity account from the first identity account sequence according to the matching result.

The probability condition comprises a preset condition and a condition determined according to the sequencing condition of the first identity account sequence. Optionally, the probability condition includes at least one of a probability numerical condition and a probability threshold condition.

In an optional embodiment, when the account recommendation method is used for recommending the first identity account, the account recommendation method may be summarized as the following three processes, which are respectively: (one) merchant preferences; (II) flow rate is preferable; and (III) performing joint modeling.

Online-Offline integration (O2O, Online To Offline), refers To combining Offline business opportunities with the internet, making the internet the foreground of Offline transactions. The concept of O2O is quite broad and illustrative, and when a chain of industries involves both on-line and off-line, it is commonly referred to as O2O.

Alternatively, in a network environment, a connection is often established between a merchant and a user through an advertisement, the merchant attracts the user through the advertisement, the user selects the merchant according to the advertisement, and the like, and the advertisement is generally implemented in a bid advertisement and a contract advertisement form.

The bidding advertisement is an advertisement form which is independently released and managed by a user, and merchants perform price ranking by adjusting prices and compete in a mode of reducing prices or raising prices. In the bidding advertising mode, traffic distribution is mainly based on factors such as estimated click rate, estimated deal rate and estimated bid, so that some merchants may not obtain stable advertising exposure opportunities.

Contract advertisement is a business model based on contracts, which means that media and advertisers (merchants) agree to deliver advertisements of the advertisers (merchants) on certain advertisement positions in a fixed way in a certain time period, usually, the total amount of the needs of the advertisers and the price of orders are explicitly written in the contracts, and the platform makes a guarantee delivery according to the contracts, namely: the delivery quantity of the advertisements is ensured to meet the requirements of the advertisers. Contract advertisements can provide advertisers with an effective and consistent advertisement exposure opportunity to reach users versus bid advertisements.

In an optional embodiment, the performance rate of contracts and the personalized matching situation of the flow and the advertisement are comprehensively measured, and the account number recommendation method is realized through a merchant preference method, a flow preference method and a joint modeling method.

(one) merchant preferences

In an optional embodiment, contracts corresponding to a plurality of merchants are obtained, the merchants corresponding to the contracts are internally sorted according to the basic situation of the contracts, and the merchants of the contracts to be released are selected.

The basic case of the contract is used to indicate contract information described in the contract, such as: c, marking a contract corresponding to the merchant C to be delivered to the contract for the user aged 18-35, determining the contract as contract information, and when the user aged 18-35, taking the merchant C as one of merchants to be subjected to internal sorting; or, if the contract type corresponding to the merchant C is a food contract, determining the classification as contract information, and when the user views the food list, taking the merchant C as one of merchants to be subjected to internal ranking, and the like.

Optionally, the ranking formula for ranking the merchants considers the merchant contract service rate and also considers the click rate, the conversion rate, the guest unit price and other pre-estimated values of the merchant, so as to improve the accuracy of traffic delivery. Illustratively, the ranking formula for ranking the merchants is shown as the ranking formula in step 220 above.

(II) flow rate optimization

In an optional embodiment, the preset traffic delivery probability is adjusted according to the estimated traffic data and the historical traffic data of different merchant contract advertisements delivered in the merchant ordering situation.

Optionally, when the estimated flow data is obtained, the estimation mode adopted is a real-time estimation mode, that is: and estimating the flow data in the future time period in real time according to the historical flow data, updating the historical flow data along with the time, and continuing the prediction process of the flow data in the future time period through the updated historical flow data.

In an alternative embodiment, the process of updating the preset traffic delivery probability based on the historical traffic data and the predicted traffic data may be implemented by using the adjustment formula in step 240.

(III) Joint modeling

The merchant is preferably used for deciding about a recommended merchant (delivered contract advertisement), and the traffic is preferably used for deciding whether to recommend the merchant to the user (deliver the contract advertisement to the user) in the recommendation process.

Illustratively, after the two stages of merchant optimization and flow optimization, the two stages are subjected to linkage optimal solution under the same optimization target, so that a better delivery effect is achieved, and a high-efficiency and high-quality delivery process of contract advertisements is realized.

In an alternative embodiment, as shown in FIG. 4, the three phases are referred to as a federated optimization process, wherein a plurality of contracts (contract 1, contract 2, contract 3 … …) are included in contract column 410, the contracts indicating the merchants as being contract advertisements determined by the platform based on which the platform recommends the merchants to the user. Illustratively, different contracts correspond to different merchants.

Optionally, internally ranking the merchants based on a merchant preference (internal ranking) 420 process, and determining candidate delivery merchants 430; adjusting a preset traffic distribution probability based on a traffic optimization (delivery service rate) 440 process, determining whether to deliver traffic based on the adjusted traffic distribution probability, and determining a delivery result 450; placement results 450 determined based on the traffic preference (placement service rate) 440 process are jointly analyzed with candidate placement merchants 430 determined based on the merchant preference (internal ranking) 420 process, ultimately placing advertisements to the user.

In summary, historical traffic data of the first identity accounts in a historical time period is acquired, recommendation degrees of the first identity accounts are sorted based on the historical traffic data, and a target recommendation probability capable of better coordinating relationships among different first identity accounts is determined. The process of recommending the first identity account to the second identity account more flexibly is achieved by means of the target recommendation probability, the historical flow data of one first identity account is better, other better first identity accounts are ignored, and therefore the recommended first identity account is matched with the requirements of the second identity account while the fact that enough first identity accounts are recommended to the second identity account is guaranteed by means of the target recommendation probability, invalid account recommendation is avoided, and the process of recommending the first identity account to the second identity account with high efficiency and high quality is achieved.

In the embodiment of the application, the account recommendation method is realized in a merchant optimization, flow optimization and combined modeling mode, after merchants are sequenced by adopting a sequencing formula, a contract service rate is adjusted by adopting an adjustment formula based on parameters such as historical flow data and estimated flow data of the merchants, and a target recommendation probability capable of better coordinating the relationship among different first identity accounts is obtained, so that advertisements and flow are matched in a personalized manner, on the basis of ensuring advertisement guarantee delivery, computing resources and network resources consumed in an invalid pushing process are reduced, the requirements of the recommended first identity account and the recommended second identity account are matched better, invalid account recommendation is avoided, the matching effect of advertisements of users and merchants is enhanced, and the experience feeling of the users and the merchants is improved.

In an optional embodiment, the account recommendation method is applied to a take-away scene. Illustratively, as shown in fig. 5, the embodiment shown in fig. 3 can also be implemented as the following steps 510 to 550.

Take-out is a typical O2O platform, and advertisers (take-out merchants) usually sign contracts with the platform to perform the process of delivering take-out advertisements with the contracted advertisements in order to obtain effective and stable advertisement exposure opportunities to reach users.

Optionally, when the platform performs the delivery process of the takeaway advertisement through the contract, if the delivery effect is ignored while paying attention to the delivery amount of the advertisement, the reissue probability of the merchant making the contract with the platform is easily reduced, and the consumption times of the user using the platform are also easily reduced. Based on the consideration of the commercial tenant and the user, the method for recommending the commercial tenant is optimized, schematically, in an outsourcing scene, the platform provides a selling channel for the commercial tenant, and the platform provides a purchasing channel for the user, so that the association between the commercial tenant and the user is realized.

At step 510, historical traffic data of at least one takeaway merchant account over a historical period of time is obtained.

Optionally, at least one take-away merchant account is determined from the plurality of merchant accounts based on different contract advertisements. For example: if the X commercial tenant provides the takeout distribution service, acquiring historical flow data of an X commercial tenant account corresponding to the X commercial tenant; and if the Y merchant does not provide the takeout distribution service, historical flow data of the Y merchant is not acquired, and the like.

Illustratively, the historical traffic data of the take-away merchant account includes historical click-through rate data, historical conversion rate data, historical customer unit price data, and the like. The historical time period is used to indicate a time period before the current time, such as: past week, past month, etc.

Step 520, based on the historical traffic data, predicting to obtain estimated traffic data of at least one take-away merchant account in a future time period.

Illustratively, after historical traffic data of a plurality of take-away merchant accounts supporting take-away delivery services is obtained, traffic data of each take-away merchant account in a future time period is predicted. For example: predicting the flow data of the X merchant account in the next month based on the historical flow data of the X merchant account in the past month, and determining the predicted flow data of the X merchant account in the next month; or predicting the flow data of the X merchant account in the future week based on the past one-month historical flow data of the X merchant account, and determining the predicted flow data of the X merchant account in the future week; or predicting the traffic data of the X merchant account in the future day based on the historical traffic data of the X merchant account in the past week, and determining the estimated traffic data of the X merchant account in the future day and the like.

Step 530, in response to receiving a takeaway merchant recommendation request sent by a user account, sorting the takeaway merchant accounts by using a first recommendation algorithm through historical traffic data to obtain a takeaway merchant account sequence.

Illustratively, a channel for providing a sales service for a take-away merchant account is an application program, the take-away merchant account can provide services such as a menu and an order receiving through the application program, and a user can implement operations such as order placing and order returning through the application program. Optionally, the manner in which the user account sends the takeaway merchant recommendation request includes: triggering a takeout control in the application program; or open the application, etc.

Wherein the contract service rate is used to indicate a probability of recommending a takeaway merchant account to the user account. Illustratively, after the estimated flow data is obtained, the take-away merchant account numbers are sorted according to different estimated flow data corresponding to different merchants and different contract service rates corresponding to different merchants, so as to obtain a take-away merchant account number sequence composed of a plurality of take-away merchant account numbers.

And 540, determining a target recommendation probability of at least one merchant by adopting a second recommendation algorithm based on the historical flow data.

Optionally, in consideration of flexibility when recommending the take-away merchant account to the user account, the contract service rate is adjusted, and schematically, a preset adjustment coefficient is predetermined, and the contract service rate is adjusted by the preset adjustment coefficient, the historical flow data and the estimated flow data, so that the target recommendation probability is determined.

Step 550, determining a takeaway merchant account recommended to the user account from the sequence of takeaway merchant accounts based on the target recommendation probability.

Optionally, after the target recommendation probability is determined, determining a condition of the takeaway merchant account recommended to the user account according to the numerical value of the target recommendation probability. The target recommendation probability is used for indicating the probability of recommending the take-away merchant account to the user account, which is obtained after the contract service rate is adjusted.

Illustratively, the larger the numerical value of the target recommendation probability is, the larger the probability of recommending the take-out merchant account to the user account is; the smaller the value of the target recommendation probability, the smaller the probability of recommending the take-away merchant account to the user account. Optionally, when the contract service rate is adjusted, the target recommendation probability of the takeaway merchant account with a larger value in the contract service rate is controlled not to infinitely approach 1, and the target recommendation probability of the takeaway merchant account with a smaller value in the contract service rate is controlled not to infinitely approach 0, that is: the relationship between the accounts of different takeaway merchants is balanced as much as possible, and not only are better historical traffic data omitted, but also better new stores and the like are omitted.

In the embodiment of the application, the account number recommendation method is applied to a takeout scene, after historical flow data of different takeout merchants are acquired, estimated flow data of a future time period is obtained, the different takeout merchants are sequenced, and different contract service rates corresponding to the different takeout merchants are adjusted by adopting an adjustment formula based on parameters such as the historical flow data and the estimated flow data corresponding to the different takeout merchants to obtain a target recommendation probability, so that the relationship between the different takeout merchants can be better coordinated, the recommended takeout merchants are more matched with the requirements of users, invalid account number recommendation is avoided, and meanwhile, the human-computer interaction efficiency of the users is improved.

Fig. 6 is a block diagram of an account recommending apparatus according to an exemplary embodiment of the present application, and as shown in fig. 5, the apparatus includes the following components:

an obtaining module 610, configured to obtain historical traffic data of at least one first identity account in a historical time period, where the first identity account is used to provide a target service;

a sorting module 620, configured to, in response to receiving an account recommendation request sent by a second identity account, sort recommendation degrees of the first identity account through the historical traffic data by using a first recommendation algorithm to obtain a first identity account sequence, where the first recommendation algorithm is used to sort the first identity account by using a contract service rate as a sorting core, and the second identity account is used to receive the target service provided by the first identity account;

a determining module 630, configured to determine a target recommendation probability of the at least one first identity account based on the historical traffic data by using a second recommendation algorithm, where the second recommendation algorithm is configured to perform probability prediction on the target recommendation probability for the first identity account with recommendation validity as a core;

a recommending module 640, configured to determine, from the sequence of first identity accounts, a first identity account recommended to the second identity account based on the target recommendation probability.

In an optional embodiment, the recommending module 640 is further configured to intercept the first k first identity account numbers from the first identity account number sequence to obtain a sequence segment, where k is a positive integer; and determining a first identity account number recommended to the second identity account number from the sequence segments based on the target recommendation probability.

In an optional embodiment, the recommending module 640 is further configured to match the target recommending probability of the first identity account with a probability condition to obtain a matching result, where the matching result includes a matching success result; and determining a first identity account number which accords with the matching success result from the sequence fragment as a first identity account number recommended to the second identity account number.

In an optional embodiment, the second recommendation algorithm includes a preset adjustment coefficient;

the determining module 630 is further configured to perform an average operation on the historical traffic data of the at least one first identity account to obtain historical average data; analyzing the historical average data and estimated flow data of the ith first identity account in a future time period based on the preset adjusting coefficient to obtain the target recommendation probability of the ith first identity account, wherein the estimated flow data is flow data obtained by predicting the ith first identity account in the future time period, and i is a positive integer.

In an optional embodiment, the preset adjustment coefficient includes a first preset coefficient and a second preset coefficient;

the determining module 630 is further configured to determine a product of the first preset coefficient and the historical average data to obtain first data; determining the product of the second preset coefficient and the estimated flow data corresponding to the ith first identity account to obtain second data; and adjusting the contract service rate of the ith first identity account according to the value obtained by adding the first data and the second data, and determining the target recommendation probability of the ith first identity account.

As shown in fig. 7, in an optional embodiment, the sorting module 620 further includes:

an adjusting unit 621, configured to adjust, based on the historical traffic data, estimated traffic data corresponding to each first identity account in the at least one first identity account by using the first recommendation algorithm, and determine an adjustment result corresponding to each first identity account, where the estimated traffic data is traffic data predicted for the first identity account in a future time period;

a determining unit 622, configured to perform recommendation ranking on the first identity accounts based on an adjustment result corresponding to each first identity account, and determine the first identity account sequence.

In an optional embodiment, the adjusting unit 621 is further configured to determine an adjusting parameter for adjusting the predicted traffic data based on the historical traffic data; and adjusting the estimated flow data corresponding to each first identity account according to the adjustment parameters and the contract service rate corresponding to each first identity account in the at least one first identity account, and determining an adjustment result corresponding to each first identity account.

In an optional embodiment, the apparatus further comprises:

the prediction module 650 is configured to obtain first historical traffic data of a first historical time period and second historical traffic data of a second historical time period in the historical traffic data; predicting the flow data of the first historical time period by using a candidate prediction model to obtain predicted flow data corresponding to the first historical flow data, wherein the candidate prediction model is a model to be trained; training the candidate prediction model based on the first historical flow data and the predicted flow data to obtain a data prediction model; and inputting second historical flow data corresponding to the at least one first identity account into the data prediction model to obtain the predicted flow data.

In an optional embodiment, the prediction module 650 is further configured to determine a loss value based on the first historical flow data and the predicted flow data for the at least one first identity account for the first historical period of time; training the candidate pre-estimation model by using the loss value; and responding to the training of the candidate prediction model to achieve a training effect, and obtaining the data prediction model.

In summary, historical traffic data of the plurality of first identity accounts in a historical time period is obtained, a first recommendation algorithm with a contract service rate as a ranking core is used for ranking recommendation degrees of the plurality of first identity accounts, and a second recommendation algorithm with recommendation effectiveness as a core is used for determining a target recommendation probability based on the historical traffic data. The recommendation probability is analyzed under double-layer parallel conditions of contract service rate and recommendation effectiveness, the recommendation accuracy is improved, computing resources and network resources consumed in an invalid pushing process are reduced, the first identity account numbers with enough quantity are recommended to the second identity account numbers, the requirements of the recommended first identity account numbers and the second identity account numbers are matched better, invalid account number recommendation is avoided, and meanwhile the human-computer interaction efficiency of a user is improved.

It should be noted that: the account recommending device provided in the above embodiment is only illustrated by dividing the functional modules, and in practical applications, the functions may be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the account recommending device and the account recommending method provided by the above embodiments belong to the same concept, and specific implementation processes thereof are detailed in the method embodiments and are not described herein again.

Fig. 8 shows a schematic structural diagram of a server according to an exemplary embodiment of the present application. The server 800 includes a Central Processing Unit (CPU) 801, a system Memory 804 including a Random Access Memory (RAM) 802 and a Read Only Memory (ROM) 803, and a system bus 805 connecting the system Memory 804 and the CPU 801. The server 800 also includes a mass storage device 806 for storing an operating system 813, application programs 814, and other program modules 815.

The mass storage device 806 is connected to the central processing unit 801 through a mass storage controller (not shown) connected to the system bus 805. The mass storage device 806 and its associated computer-readable media provide non-volatile storage for the server 800. That is, the mass storage device 806 may include a computer-readable medium (not shown) such as a hard disk or Compact disk Read Only Memory (CD-ROM) drive.

Without loss of generality, computer readable media may comprise computer storage media and communication media. Computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes RAM, ROM, Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), flash Memory or other solid state Memory technology, CD-ROM, Digital Versatile Disks (DVD), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage, or other magnetic storage devices. Of course, those skilled in the art will appreciate that computer storage media is not limited to the foregoing. The system memory 804 and mass storage device 806 as described above may be collectively referred to as memory.

According to various embodiments of the present application, server 800 may also operate as a remote computer connected to a network through a network, such as the Internet. That is, the server 800 may be connected to the network 812 through the network interface unit 811 coupled to the system bus 805, or may be connected to other types of networks or remote computer systems (not shown) using the network interface unit 811.

The memory further includes one or more programs, and the one or more programs are stored in the memory and configured to be executed by the CPU.

An embodiment of the present application further provides a computer device, where the computer device includes a processor and a memory, where the memory stores at least one instruction, at least one program, a code set, or a set of instructions, and the at least one instruction, the at least one program, the code set, or the set of instructions is loaded and executed by the processor to implement the account recommendation method provided in each method embodiment.

An embodiment of the present application further provides a computer-readable storage medium, where at least one instruction, at least one program, a code set, or an instruction set is stored on the computer-readable storage medium, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor, so as to implement the account recommendation method provided by each method embodiment.

Embodiments of the present application also provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, so that the computer device executes the account recommendation method in any one of the above embodiments.

Optionally, the computer-readable storage medium may include: a Read Only Memory (ROM), a Random Access Memory (RAM), a Solid State Drive (SSD), or an optical disc. The Random Access Memory may include a resistive Random Access Memory (ReRAM) and a Dynamic Random Access Memory (DRAM). The above-mentioned serial numbers of the embodiments of the present application are merely for description and do not represent the merits of the embodiments.

It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

The above description is only exemplary of the present application and should not be taken as limiting, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An account recommendation method is characterized by comprising the following steps:

2. The method of claim 1, wherein determining the first identity account from the sequence of first identity accounts to recommend to the second identity account based on the target recommendation probability comprises:

intercepting the first k first identity account numbers from the first identity account number sequence to obtain a sequence segment, wherein k is a positive integer;

determining, from the sequence segments, a first identity account recommended to the second identity account based on the target recommendation probability.

3. The method of claim 2, wherein determining, from the sequence segments, a first identity account recommended to the second identity account based on the target recommendation probability comprises:

matching the target recommendation probability of the first identity account with probability conditions to obtain a matching result, wherein the matching result comprises a matching success result;

and determining a first identity account number which accords with the matching success result from the sequence fragment as a first identity account number recommended to the second identity account number.

4. The method according to claim 1, wherein the second recommendation algorithm comprises a preset adjustment coefficient;

determining, by using a second recommendation algorithm, a target recommendation probability for the at least one first identity account based on the historical traffic data, including:

performing mean value operation on the historical flow data of the at least one first identity account to obtain historical average data;

analyzing the historical average data and estimated flow data of the ith first identity account in a future time period based on the preset adjusting coefficient to obtain the target recommendation probability of the ith first identity account, wherein the estimated flow data is flow data obtained by predicting the ith first identity account in the future time period, and i is a positive integer.

5. The method of claim 4, wherein the preset adjustment coefficients comprise a first preset coefficient and a second preset coefficient;

analyzing the historical average data and the estimated flow data of the ith first identity account in a future time period based on the preset adjustment coefficient to obtain the target recommendation probability of the ith first identity account, including:

determining the product of the first preset coefficient and the historical average data to obtain first data;

determining the product of the second preset coefficient and the estimated flow data corresponding to the ith first identity account to obtain second data;

and adjusting the contract service rate of the ith first identity account according to the value obtained by adding the first data and the second data, and determining the target recommendation probability of the ith first identity account.

6. The method according to any one of claims 1 to 3, wherein the obtaining a first sequence of identity accounts by ranking recommendation degrees of the first identity accounts through the historical traffic data by using a first recommendation algorithm comprises:

based on the historical flow data, the estimated flow data corresponding to each first identity account in the at least one first identity account is adjusted by adopting the first recommendation algorithm, and an adjustment result corresponding to each first identity account is determined, wherein the estimated flow data are flow data obtained by predicting the first identity account in a future time period;

and performing recommendation degree sequencing on the first identity accounts based on the adjustment result corresponding to each first identity account, and determining the sequence of the first identity accounts.

7. The method of claim 6, wherein the adjusting the estimated flow data corresponding to each of the at least one first identity account by using the first recommendation algorithm based on the historical flow data to determine an adjustment result corresponding to each first identity account comprises:

determining an adjustment parameter for adjusting the pre-estimated flow data based on the historical flow data;

and adjusting the estimated flow data corresponding to each first identity account according to the adjustment parameters and the contract service rate corresponding to each first identity account in the at least one first identity account, and determining an adjustment result corresponding to each first identity account.

8. The method of claim 7, wherein after obtaining historical traffic data for the at least one first identity account over a historical period of time, further comprising:

acquiring first historical flow data of a first historical time period and second historical flow data of a second historical time period in the historical flow data;

predicting the flow data of the first historical time period by using a candidate prediction model to obtain predicted flow data corresponding to the first historical flow data, wherein the candidate prediction model is a model to be trained;

training the candidate prediction model based on the first historical flow data and the predicted flow data to obtain a data prediction model;

and inputting second historical flow data corresponding to the at least one first identity account into the data prediction model to obtain the predicted flow data.

9. The method of claim 8, wherein training the candidate predictive model based on the first historical flow data and the predicted flow data to obtain a data predictive model comprises:

determining a loss value based on first historical traffic data and the predicted traffic data for the at least one first identity account over the first historical period of time;

training the candidate pre-estimation model by using the loss value;

and responding to the training of the candidate prediction model to achieve a training effect, and obtaining the data prediction model.

10. The method according to any one of claims 1 to 3,

the historical traffic data includes at least one of click through rate data, conversion rate data, and customer price data.

11. An account recommending apparatus, comprising:

12. A computer device comprising a processor and a memory, the memory having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, the at least one instruction, the at least one program, the set of codes, or the set of instructions being loaded and executed by the processor to implement the account recommendation method of any one of claims 1 to 10.

13. A computer-readable storage medium having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by a processor to implement the account recommendation method of any one of claims 1 to 10.

14. A computer program product comprising a computer program or instructions which, when executed by a processor, carries out the account recommendation method of any one of claims 1 to 10.