CN108665329B - Commodity recommendation method based on user browsing behavior - Google Patents
Commodity recommendation method based on user browsing behavior Download PDFInfo
- Publication number
- CN108665329B CN108665329B CN201710200470.6A CN201710200470A CN108665329B CN 108665329 B CN108665329 B CN 108665329B CN 201710200470 A CN201710200470 A CN 201710200470A CN 108665329 B CN108665329 B CN 108665329B
- Authority
- CN
- China
- Prior art keywords
- user
- commodity
- users
- vector
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000004364 calculation method Methods 0.000 claims abstract description 39
- 238000012549 training Methods 0.000 claims description 66
- 238000012821 model calculation Methods 0.000 claims description 15
- 238000012937 correction Methods 0.000 claims description 10
- 238000007635 classification algorithm Methods 0.000 claims description 7
- 230000004044 response Effects 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 2
- 230000006399 behavior Effects 0.000 abstract description 51
- 238000012545 processing Methods 0.000 description 13
- 230000000875 corresponding effect Effects 0.000 description 8
- 238000010606 normalization Methods 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000000605 extraction Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000003292 diminished effect Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a commodity recommendation method based on user browsing behaviors, which comprises the following steps: analyzing, by the purchase behavior predictor, merchandise browsing history data of one or more users, determining a first set of users that are more likely to make purchases over a future period of time; for each user in the determined first set of users: acquiring one or more commodities browsed by the user from commodity browsing history data of the user, performing commodity similarity calculation on the one or more commodities and target commodities, determining that the user is more likely to purchase the target commodities in the future period based on the commodity similarity calculation result, and adding the user to a second user set; and recommending the target commodity to each user in the determined second set of users.
Description
Technical Field
The invention relates to a commodity recommendation method, a commodity recommendation system, electronic equipment and a readable storage medium based on user browsing behaviors.
Background
With the rise of the internet, more and more people go to shopping on the internet. The competition of the e-commerce industry is also becoming more and more intense. Various marketing approaches are also layered. In order to promote sales of commodities, marketing departments will promote the sales of commodities, and various marketing information is usually sent to users through emails, short messages, mobile phone push messages and the like. Mining potential users who are most likely to purchase goods has become an integral part of the marketing department.
Before sending marketing information to users, it is necessary to analyze users who potentially purchase the targeted merchandise being promoted. At present, three common commodity recommendation methods are generally available:
Recommendation based on crowd labels: various tags of the user, such as gender tags, income tags, consumer tags, occupation tags, etc., are analyzed for marketing to the crowd having the corresponding tag. Based on the crowd labeling method, the actual effect is not good, the identification of the labels is a complex process, and many labels are a dynamic change process, so that the accuracy of the labels is not high.
Recommendation based on commodities: finding out commodities similar to the target marketing commodity, and then extracting the user purchasing the similar commodity for marketing. The commodity-based method ignores the characteristics of the user, such as the fact that two commodities are very similar but different brands, and if the brand loyalty of the user is very high, the marketing can even cause the user to feel disliked.
Recommendation based on model training: the recommendation based on model training currently used mainly uses user evaluation and scoring of commodities as basic data. However, user ratings and ratings are affected by many subjective factors and data is not fully collected in the event that the user is not interested in the ratings or ratings and does not perform the corresponding action. Therefore, model training is performed by evaluation or scoring, the most realistic purchasing habit of the user cannot be obtained, and the crowd of potential purchasing target commodities cannot be accurately analyzed.
Disclosure of Invention
Therefore, the embodiment of the invention provides a commodity recommending method based on user browsing behaviors, which can accurately analyze the crowd purchasing potential target commodities by using a method of combining model calculation with similarity calculation, and improves the actual order conversion rate of users. Analyzing the historical data of each user by using model calculation to determine the possibility of purchasing each user in a future period, so as to obtain a user set which is more likely to perform purchasing behavior; and then, carrying out similarity calculation on the target commodity and the commodity browsed by the user who is likely to carry out purchasing behavior, and finally obtaining the user who is likely to purchase the target commodity in a future period. Thereby enabling the recommendation of the target commodity to a user who is likely to purchase the target commodity.
In order to achieve the above object, according to an aspect of the present invention, there is provided a commodity recommendation method based on browsing behavior of a user.
According to one aspect of the technical scheme, the commodity recommendation method based on the browsing behavior of the user comprises the following steps: analyzing, by the purchase behavior predictor, merchandise browsing history data of one or more users, determining a first set of users that are more likely to make purchases over a future period of time; for each user in the determined first set of users: acquiring one or more commodities browsed by the user from commodity browsing history data of the user, performing commodity similarity calculation on the one or more commodities and target commodities, determining that the user is more likely to purchase the target commodities in the future period based on the commodity similarity calculation result, and adding the user to a second user set; and recommending the target commodity to each user in the determined second set of users.
Optionally, wherein analyzing, by the purchase behavior predictor, the merchandise browsing history data of the one or more users, determining the first set of users more likely to make purchases during the future period comprises: for each of the one or more users: determining, by the purchase behavior predictor, a likelihood that the user will conduct a purchase behavior during the future period based on the merchandise browsing history data of the user; in response to determining that the likelihood is greater than a threshold, the user is added to the first set of users.
Optionally, training the purchase behavior predictor, wherein the training comprises: receiving sample commodity browsing history data of a plurality of specific users; generating training data associated with the plurality of particular users based on the received sample merchandise browsing history data; training the purchase behavior predictor using the training data.
Optionally, generating training data associated with the plurality of particular users includes: for each specific user in the plurality of specific users, extracting characteristics of sample basic historical data of the specific user based on predefined user behavior characteristics and adding labels to generate training data associated with the specific user; training data associated with each particular user is aggregated, the aggregated training data is normalized, and the training data associated with the plurality of particular users is generated.
Optionally, wherein performing the commodity similarity calculation on the one or more commodities to the target commodity includes: constructing a target commodity vector of the target commodity based on a predefined commodity attribute feature model aiming at the class of the target commodity; for each user in the first set of users: based on the predefined commodity attribute feature model, constructing a comparison commodity vector of commodities in the same category as the target commodity in commodities browsed by the user, wherein the target commodity vector and the comparison commodity vector contain the same vector dimension; and carrying out similarity calculation on the target commodity vector and the comparison commodity vector.
Optionally, wherein the training data of the user is represented in a vector.
Alternatively, the method in which the model is trained is a classification algorithm.
Optionally, wherein the similarity calculation is a cosine similarity calculation with a correction coefficient.
According to another aspect of the technical scheme of the invention, a commodity recommendation device based on user browsing behaviors is provided.
According to another aspect of the present invention, an apparatus for recommending goods based on browsing behavior of a user includes: the model calculation module is used for: analyzing, by the purchase behavior predictor, merchandise browsing history data of one or more users, determining a first set of users that are more likely to make purchases over a future period of time; for each user in the determined first set of users: acquiring one or more commodities browsed by the user from commodity browsing history data of the user, performing commodity similarity calculation on the one or more commodities and target commodities, determining that the user is more likely to purchase the target commodities in the future period based on the commodity similarity calculation result, and adding the user to a second user set; and recommending the target commodity to each user in the determined second set of users.
Optionally, the model calculation module is further configured to: for each of the one or more users: determining, by the purchase behavior predictor, a likelihood that the user will conduct a purchase behavior during the future period based on the merchandise browsing history data of the user; in response to determining that the likelihood is greater than a threshold, the user is added to the first set of users.
Optionally, the model calculation module is further configured to train the purchase behavior predictor, wherein the training includes: receiving sample commodity browsing history data of a plurality of specific users; generating training data associated with the plurality of particular users based on the received sample merchandise browsing history data; training the purchase behavior predictor using the training data.
Optionally, wherein the model calculation module is further configured to: for each specific user in the plurality of specific users, extracting characteristics of sample basic historical data of the specific user based on predefined user behavior characteristics and adding labels to generate training data associated with the specific user; training data associated with each particular user is aggregated, the aggregated training data is normalized, and the training data associated with the plurality of particular users is generated.
Optionally, the similarity calculation module is further configured to: constructing a target commodity vector of the target commodity based on a predefined commodity attribute feature model aiming at the class of the target commodity; for each user in the first set of users: based on the predefined commodity attribute feature model, constructing a comparison commodity vector of commodities in the same category as the target commodity in commodities browsed by the user, wherein the target commodity vector and the comparison commodity vector contain the same vector dimension; and carrying out similarity calculation on the target commodity vector and the comparison commodity vector.
Optionally, wherein the training data of the user is represented in a vector.
Alternatively, the method in which the model is trained is a classification algorithm.
Optionally, wherein the similarity calculation is a cosine similarity calculation with a correction coefficient.
To achieve the above object, according to still another aspect of the present invention, there is provided an electronic apparatus.
An electronic apparatus of the present invention includes: one or more processors; and the storage device is used for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors realize the commodity recommendation method based on the browsing behaviors of the user.
To achieve the above object, according to still another aspect of the present invention, there is provided a computer-readable storage medium.
A computer readable storage medium of the present invention has stored thereon a computer program which, when executed by a processor, implements the method for commodity recommendation based on user browsing behavior provided by the present invention.
One embodiment of the above invention has the following advantages or benefits: because the technical means of combining model calculation with similarity calculation is adopted, the problem that a target user cannot be accurately positioned in the prior art is solved, and the technical effect of accurately analyzing potential crowds buying target commodities is achieved.
Further effects of the above-described non-conventional alternatives are described below in connection with the embodiments.
Drawings
The drawings are included to provide a better understanding of the invention and are not to be construed as unduly limiting the invention. Wherein:
FIG. 1 is a general flow chart of a merchandise recommendation method based on user browsing behavior according to an embodiment of the invention;
FIG. 2 is a flow chart of feature processing to generate a sample of a user according to an embodiment of the present invention;
FIG. 3 is a flow chart of model training to generate model files according to an embodiment of the present invention;
FIG. 4 is a flow chart of similarity calculation according to an embodiment of the present invention;
FIG. 5 is a system block diagram of a merchandise recommendation system based on user browsing behavior in accordance with an embodiment of the invention;
fig. 6 is a schematic diagram of a hardware structure of an electronic device for implementing a commodity recommendation method based on user browsing behavior according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings, in which various details of the embodiments of the present invention are included to facilitate understanding, and are to be considered merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the invention. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
The invention uses model training to screen the user for the first time based on the sample of the user after the characteristic processing, predicts the user who is likely to purchase; and then, the user is screened for the second time by comparing the target commodity with the historical commodity browsed by the user who is likely to purchase, so as to predict the user who is likely to purchase the target commodity.
Sample of user: and performing feature processing and labeling on the user browsing history data to generate a sample of the user. Samples of multiple users may be used as training data for model training. In an example embodiment, user Amy browses 2 items at day 3 months 1 and adds one of the items to the shopping cart, browses 4 items at day 3 months 2, and purchases 2 items at day 3 months 4. The collection of these actions constitutes the underlying historical data of the user Amy. Processing the basic history data of Amy through a feature processing step, and finally forming a sample with features and labels. In one embodiment, the tag indicates whether the user purchased the merchandise. For example, 0 indicates that no merchandise has been purchased and 1 indicates that merchandise has been purchased. In alternative embodiments, the tag may represent other information, such as 0 indicating that no more than a predetermined number of items were purchased, and 1 indicating that more than a predetermined number of items were purchased.
The characteristics are as follows: we mine the features of various user behaviors through the history of the user's browsing. The features are defined according to a predetermined feature definition rule. In one embodiment, the characteristic values are represented by numbers.
And (3) model: models are generated by model training of thousands of samples. For example, a model is calculated by training samples of thousands of users. In one embodiment, the model may be a function. In an alternative embodiment, the model may be an algorithm. A model may be applied to each user's sample to predict the user's behavior in future time periods.
Model file: the model file holds the parameters of the model. And (3) carrying out mathematical operation on the user sample by introducing parameters in the model file into the model, and finally outputting a probability value between 0 and 1.
Fig. 1 is a general flow chart of a commodity recommendation method based on user browsing behavior according to an embodiment of the present invention. As shown in fig. 1, fig. 1 includes steps S11-S15.
In step S11 of fig. 1, a browsing history of each of a plurality of users is subjected to feature processing, and samples of the plurality of users are generated as training data. See fig. 2 for detailed steps.
Turning to fig. 2, fig. 2 is a flow chart illustrating a process of feature processing and sample generation based on historical data of a user, according to an embodiment of the present invention. In step S11-1 of FIG. 2, base history data for one or more users browsing websites is collected. In one embodiment, the underlying history data is from a business data store such as a flow table, click table, order table, reservation table, etc. stored in a storage unit of the distributed file system. Techniques such as mapreduce+hive may be employed to capture merchandise characteristics. In one embodiment, data may be processed in a user dimension, and the granularity of user browsing may be collected in a Stock Keeping Unit (SKU) number. In one embodiment, a timer is used to set the data acquisition time to achieve automatic data acquisition. For example, the night system is loaded less, and the timer can be set to automatically start data acquisition at night to avoid affecting the main business.
In step S11-2 of FIG. 2, feature extraction is performed on the base data for each user, and a user sample vector is generated. Table 1 below exemplarily lists example feature definition rules for user behavior. The feature extraction process screens dimensions and features such as those listed in table 1 according to predefined feature rules. In an example embodiment, for example, where the user Amy has focused on 4 items for the last week, but has not performed an action to join the shopping cart, the feature extraction process sets the "number of items focused" feature to 4 and the "join shopping cart" feature to 0. The desire of users to purchase goods is gradually diminished over time. Through analysis of the historical purchase records of the user, the closer the browsing time is, the greatest influence is on the commodity purchased by the user. In one embodiment, for exemplary purposes only, the feature extraction time dimension may be defined as the last 7 days, with example dimensions for feature extraction and corresponding example feature definitions shown in table 1. In other embodiments different features and extraction times may be defined.
In step S11-3 of FIG. 2, the feature is cleaned of outlier data and normalized for each dimension of the feature. Since the magnitude of each feature is not uniform, all dimensions of the user's sample vector must be normalized. In an exemplary embodiment, the number of commodities browsed by the user Amy is 5, the number of days between the last purchase and the present time is 200, and normalization processing is performed on '5' and '200' for processing by the algorithm. In one embodiment, the Min-Max Normalization normalization algorithm may be used to normalize the values to the interval Table 1 example feature definition rules
[0,1 ]. Alternatively, various other normalization algorithms such as Z-score may be employed. In an example embodiment, the sample of user Amy, bob, cindy, dora has features in two dimensions, "number of items of interest" and "number of items collected", respectively:
Table 2 example user samples
| User name | Concern the number of commodities | Number of stored goods |
| Amy | 2 | 4 |
| Bob | 8 | 3 |
| Cindy | 10 | 3 |
| Dora | 10000 | 2 |
In this example embodiment, the sample "number of commodities of interest" value of the user Dora is 10000, which is much larger than the values of other samples, and the Dora sample is removed as an outlier.
In one embodiment, taking the Min-Max Normalization normalization algorithm as an example, the feature value for each dimension is divided by the maximum value for that dimension feature. The user's sample is converted into the following form:
Table 3 example user samples feature cleaned
| User name | Concern the number of commodities | Number of stored goods |
| Amy | 0.2 | 1 |
| Bob | 0.8 | 0.75 |
| Cindy | 1 | 0.75 |
In step S11-4 of fig. 2, samples of the plurality of users subjected to the feature processing are output for use in the following steps. After the characterization process, the samples are labeled. In an example embodiment, the labels of the samples are defined as, for example, "positive and negative samples", and the selection criteria for the positive and negative samples are defined by parameter settings. For example, a positive sample may be defined as a user purchasing a good, and a negative sample may be defined as a user having browsing activity, but not purchasing a good. The output user sample is the sample with the feature + label. The feature-processed and labeled user sample is the basis for the subsequent operation of the present invention, which is based on the feature-processed and labeled user sample in the subsequent steps.
Returning to fig. 1, in step S12 of fig. 1, model training is performed using samples of a plurality of specific users according to predefined parameter settings, a purchase behavior predictor is generated, and a model file is output. The predefined parameters may set the start-stop time of the sample, specification of the model, sample partitioning criteria, etc. See fig. 3 for detailed steps.
Fig. 3 is a flow chart according to an embodiment of the invention. In step S12-1 of FIG. 3, the training set and the test set are partitioned. The training set is a sample set for model training, and the test set is a sample set for verifying the accuracy of the model. The 8:2 principle is generally adopted in model training. For example, 10 ten thousand samples of a particular user, 8 ten thousand as training sets and 2 ten thousand as test sets.
In step S12-2 of fig. 3, model training is performed using a specific user sample in the training set, and the model trained by the model is referred to as a purchase behavior predictor. In one embodiment, a supervised learning classification algorithm may be employed to train models (such as logistic regression, decision tree, etc. algorithms). The input data for model training is a two-dimensional matrix of user samples, where each row is a vector of user samples and each column represents a feature representing the value of the sample corresponding to the feature. For each sample, there is a model-trained target value. In embodiments employing supervised learning of the classification algorithm to train the model, there are only two target values for the user sample used for model training.
After model training on the training test set, the accuracy of the model is verified using the previously partitioned test set.
In step S12-3 of FIG. 3, the model file is output. The model file is a parameter of the purchase behavior predictor. Different model files may be generated for different scenarios. Different model files are stored in the storage unit for future selection for different predicted scenarios.
Returning to fig. 1, in step S13 of fig. 1, a sample of each user is predicted with the outputted model file to determine a first set of users likely to make purchases over a given future period of time. When the target user is predicted, a sample of the user is obtained, and then the parameters in the selected model file are utilized to carry out mathematical operation of a classification algorithm through a purchase behavior predictor, so that the purchase probability is obtained. In one embodiment, if the resulting probability is 0.8, a probability of 80% would be likely to make a purchase and a probability of 20% would not be. The purchase probability of each user is compared to a threshold, and a user having a purchase probability greater than the threshold is selected as the first set of users who are likely to make purchases.
In step S14 of fig. 1, the target commodity is compared with the commodities that each user in the first set of users who are more likely to purchase has browsed, and a second set of users who are likely to purchase the target commodity is determined. See fig. 4 for detailed steps.
Fig. 4 is a flowchart of similarity calculation according to an embodiment of the present invention. In S14-1 of fig. 4, a target commodity vector of the target commodity is constructed for the category of the target commodity based on the attribute feature model of the commodity defined in advance. The attribute features of the merchandise such as price, brand, place of origin, etc.
In S14-2 of fig. 4, for the commodities in the same category as the target commodity in the commodities once browsed by each user, a comparison commodity vector of the user is constructed based on the predefined attribute feature model of the commodity, wherein the target commodity vector and the comparison commodity vector contain the same vector dimension, and the vector dimension is the attribute feature.
In S14-3 of fig. 4, a similarity calculation is performed between the target commodity vector and the user' S reference commodity vector. In one embodiment, correction factors may be used in combination with cosine similarity to calculate.
In an embodiment employing a correction factor in combination with cosine similarity, the similarity calculation formula is as follows:
Wherein sigma is a correction coefficient, V 1 is a target commodity attribute feature vector, V 2 is a commodity attribute feature vector browsed by a user, and W i is a weight vector corresponding to each attribute. Wherein V 1 and V 2 comprise the same vector dimensions. And the goods browsed by the user are in the same category as the target goods. The vector dimensions represent the attribute features of the commodity, and the same vector dimensions represent the same commodity attributes.
V1＝(v11,v12,v13........)V2＝(v21,v22,v23........)W＝(W1,W2,W3........)
In this embodiment of the present invention, in one embodiment,
Target commodity vector is preset to V 1 = (1,) No./No.)
The attribute feature values of the target commodity are all set to 1. Alternatively, the characteristic value of the price attribute may be obtained by dividing the actual price value by a normalization coefficient (e.g., 10000).
The calculation method of the comparison commodity vector V 2 browsed by the user is as follows:
each attribute feature value in the user's browsing commodity feature vector is calculated according to the following formula. Where price attributes are special attributes, and thus are calculated in a separate "price attribute" formula, other attributes, such as brands, are calculated in an "other attribute" formula.
Taking a computer as an example, the attribute and the weight of the attribute may be, for example:
and (3) a computer: price-1.0, screen size-0.3, hot spot-0.3, brand-0.6, body color-0.3, system-0.6, core number-0.3, memory-0.5.
1. The characteristic value calculation formula of the price attribute comprises the following steps:
a i is the price of the browsed commodity, b i is the browsing times of the commodity, and n is the number of the browsed commodities.
For example, the user browses 4 commodities, the prices are 5000, 4500, 6000, 3500, and the average prices are 1, 3, 1, respectively
(5000*1+4500*3+6000*1+3500*1)/6＝4667,
Then divided by the normalization coefficient (e.g., 10000) to obtain a price characteristic value 0.4667.
2. The eigenvalue calculation formula of other attributes:
for example, when the user browses 4 items, brands A, B, C, D respectively, the total browsing times are 4 times, and brand C appears 2 times, the feature value of brand C is 2/4=0.5.
And then, calculating the similarity between each feature of the target commodity and each feature of the commodity in the same category browsed by the user respectively by adopting the cosine similarity based on the weight.
When the cosine similarity is calculated, the behavior of the user needs to be increased by using a correction coefficient from the perspective of the property of the commodity. For example, the number of times the user browses the item detail page the last week, the number of times the item detail page is clicked on the record point, and the like, and the behavior of the latest time is weighted more. The cosine similarity is improved, and the result of the cosine similarity is multiplied by a correction coefficient.
Correction coefficient calculation formula= (browsing coefficient+recording point coefficient)/2
1. The method for calculating the browsing coefficient a1 comprises the following steps:
n i is the number of browsing per day
2. Recording a point coefficient a2:
n i is the number of clicks of a record, and w i is the record weight (the top 8 record are ranked by the weight).
The calculated similarity is compared with a threshold value in S14-4 of fig. 4, and the user corresponding to the similarity greater than the threshold value is determined, generating a set of potential users who are likely to purchase the target commodity. In an example embodiment, a threshold value such as 0.8 may be set, and a user whose final similarity calculation result is greater than 0.8 is extracted as a user who is likely to purchase the target commodity.
Returning to fig. 1, in step S15 of fig. 1, the target commodity is recommended to the user in the second user set that is determined to be more likely to purchase the target commodity.
FIG. 5 is a system block diagram of a merchandise recommendation system based on user browsing behavior according to an embodiment of the invention. The system is divided into a model calculation module and a similarity calculation module. The model calculation module comprises a feature processing sub-module, a model training sub-module and a model prediction sub-module. The model calculation module carries out feature processing on the historical data of each user to obtain a sample of the user, carries out model training on the samples of a plurality of users to generate a model file, predicts the purchasing behavior of the user in the future period by using the sample of the target user and the selected model file, and determines a first user set which is likely to carry out purchasing behavior; the similarity calculation module calculates, for each user in the first set of users, a similarity between the target commodity and the historical commodity that the user has browsed, thereby determining a second set of users who are likely to purchase the target commodity.
According to an embodiment of the present invention, the present invention also provides an electronic device and a readable storage medium.
The electronic device of the present invention includes: at least one processor; and a memory communicatively coupled to the at least one processor; the memory stores instructions executable by the one processor, and the instructions are executed by the at least one processor, so that the at least one processor executes the commodity recommendation method based on the browsing behavior of the user.
The non-transitory computer readable storage medium of the present invention stores computer instructions for causing the computer to execute the method for commodity recommendation based on user browsing behavior provided by the present invention.
Fig. 6 is a schematic diagram of a hardware structure of an electronic device for implementing a method for recommending goods based on browsing behavior of a user according to an embodiment of the present invention. As shown in fig. 6, the electronic device includes: one or more processors 61 and a memory 62, one processor 61 being exemplified in fig. 6. The memory 62 is a non-transitory computer readable storage medium provided by the present invention.
The electronic device of the commodity recommendation method based on the browsing behavior of the user can further comprise: an input device 63 and an output device 64.
The processor 61, the memory 62, the input means 63 and the output means 64 may be connected by a bus or otherwise, in fig. 6 by way of example.
The memory 62 is used as a non-transitory computer readable storage medium for storing non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions/modules (e.g., the model calculation module 51 and the similarity calculation module 52 shown in fig. 5) corresponding to the method of commodity recommendation based on user browsing behavior in the embodiment of the present invention. The processor 61 executes various functional applications of the server and data processing, namely, a method of realizing commodity recommendation based on user browsing behavior in the above-described method embodiment by running non-transitory software programs, instructions, and modules stored in the memory 62.
Memory 62 may include a storage program area that may store an operating system, at least one application program required for functionality, and a storage data area; the storage data area may store data created according to the use of the apparatus for commodity recommendation based on the user's browsing behavior, and the like. In addition, the memory 62 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, memory 62 may optionally include memory remotely located with respect to processor 61, which may be connected via a network to a means of merchandise recommendation based on user browsing activity. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input device 63 may receive input numeric or character information and key signal inputs related to user settings and function controls of the device that generate merchandise recommendations based on user browsing behavior. The output device 64 may include a display device such as a display screen.
The one or more modules are stored in the memory 62 and when executed by the one or more processors 61 perform the method of merchandise recommendation based on user browsing behavior in any of the method embodiments described above.
The product can execute the method provided by the embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method. Technical details not described in detail in this embodiment may be found in the methods provided in the embodiments of the present invention.
According to the technical scheme of the embodiment of the invention, potential users purchasing target commodities are accurately excavated, the target commodities are recommended to the screened potential users, and finally the purpose of improving the order conversion rate of the users is achieved.
The above embodiments do not limit the scope of the present invention. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives can occur depending upon design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of the present invention.
Claims (16)
1. A commodity recommendation method based on user browsing behavior, the method comprising:
analyzing, by the purchase behavior predictor, merchandise browsing history data of one or more users, determining a first set of users that are more likely to make purchases over a future period of time;
For each user in the determined first set of users:
One or more goods browsed by the user are obtained from the goods browsing history data of the user,
And calculating the commodity similarity between the one or more commodities and the target commodity, wherein the commodity similarity calculation comprises the following steps:
Constructing a target commodity vector of the target commodity based on a predefined commodity attribute feature model aiming at the class of the target commodity;
For each user in the first set of users:
Based on the predefined commodity attribute feature model, constructing a comparison commodity vector of commodities in the same category as the target commodity in commodities browsed by the user, wherein the target commodity vector and the comparison commodity vector contain the same vector dimension;
performing similarity calculation on the target commodity vector and the comparison commodity vector;
Determining that the user is more likely to purchase a target commodity in the future period based on the result of the commodity similarity calculation, and
Adding the user to a second set of users; and
Recommending the target commodity to each user in the determined second set of users.
2. The method of claim 1, wherein analyzing, by the purchase behavior predictor, the merchandise browsing history data of the one or more users, determining the first set of users that are more likely to make purchases during the future period comprises:
for each of the one or more users:
Determining, by the purchase behavior predictor, a likelihood that the user will conduct a purchase behavior during the future period based on the merchandise browsing history data of the user;
in response to determining that the likelihood is greater than a threshold, the user is added to the first set of users.
3. The method of claim 1, further comprising training the purchase behavior predictor, wherein the training comprises:
receiving sample commodity browsing history data of a plurality of specific users;
Generating training data associated with the plurality of particular users based on the received sample merchandise browsing history data;
Training the purchase behavior predictor using the training data.
4. The method of claim 3, wherein generating training data associated with the plurality of particular users comprises:
For each specific user in the plurality of specific users, extracting characteristics of sample basic historical data of the specific user based on predefined user behavior characteristics and adding labels to generate training data associated with the specific user;
training data associated with each particular user is aggregated, the aggregated training data is normalized, and the training data associated with the plurality of particular users is generated.
5. The method of any of claims 3-4, wherein the training data associated with the particular user is represented in a vector.
6. The method of any one of claims 3-4, wherein the method of training is a classification algorithm.
7. The method of any of claims 1-4, wherein the similarity calculation is a cosine similarity calculation with a correction coefficient.
8. An apparatus for commodity recommendation based on user browsing behavior, comprising:
the model calculation module is used for:
analyzing, by the purchase behavior predictor, merchandise browsing history data of one or more users, determining a first set of users that are more likely to make purchases over a future period of time;
For each user in the determined first set of users:
One or more goods browsed by the user are obtained from the goods browsing history data of the user,
And calculating the commodity similarity between the one or more commodities and the target commodity, wherein the commodity similarity calculation comprises the following steps:
Constructing a target commodity vector of the target commodity based on a predefined commodity attribute feature model aiming at the class of the target commodity;
For each user in the first set of users:
Based on the predefined commodity attribute feature model, constructing a comparison commodity vector of commodities in the same category as the target commodity in commodities browsed by the user, wherein the target commodity vector and the comparison commodity vector contain the same vector dimension;
performing similarity calculation on the target commodity vector and the comparison commodity vector;
Determining that the user is more likely to purchase a target commodity in the future period based on the result of the commodity similarity calculation, and
Adding the user to a second set of users; and
Recommending the target commodity to each user in the determined second set of users.
9. The apparatus of claim 8, wherein the model calculation module is further to:
for each of the one or more users:
Determining, by the purchase behavior predictor, a likelihood that the user will conduct a purchase behavior during the future period based on the merchandise browsing history data of the user;
in response to determining that the likelihood is greater than a threshold, the user is added to the first set of users.
10. The apparatus of claim 8, wherein the model calculation module is further configured to train the purchase behavior predictor, wherein the training comprises:
receiving sample commodity browsing history data of a plurality of specific users;
Generating training data associated with the plurality of particular users based on the received sample merchandise browsing history data;
Training the purchase behavior predictor using the training data.
11. The apparatus of claim 10, wherein the model calculation module is further configured to:
For each specific user in the plurality of specific users, extracting characteristics of sample basic historical data of the specific user based on predefined user behavior characteristics and adding labels to generate training data associated with the specific user;
training data associated with each particular user is aggregated, the aggregated training data is normalized, and the training data associated with the plurality of particular users is generated.
12. The apparatus of any of claims 10-11, wherein the training data associated with the particular user is represented in a vector.
13. The apparatus of any one of claims 10-11, wherein the method of training is a classification algorithm.
14. The apparatus of any of claims 8-11, wherein the similarity calculation is a cosine similarity calculation with a correction coefficient.
15. An electronic device, comprising:
One or more processors;
storage means for storing one or more programs,
When executed by the one or more processors, causes the one or more processors to implement the method of any of claims 1-7.
16. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the method according to any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710200470.6A CN108665329B (en) | 2017-03-29 | 2017-03-29 | Commodity recommendation method based on user browsing behavior |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710200470.6A CN108665329B (en) | 2017-03-29 | 2017-03-29 | Commodity recommendation method based on user browsing behavior |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108665329A CN108665329A (en) | 2018-10-16 |
| CN108665329B true CN108665329B (en) | 2024-09-24 |
Family
ID=63787066
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710200470.6A Active CN108665329B (en) | 2017-03-29 | 2017-03-29 | Commodity recommendation method based on user browsing behavior |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108665329B (en) |
Families Citing this family (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111125503B (en) * | 2018-10-31 | 2023-04-07 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN109242649A (en) * | 2018-10-31 | 2019-01-18 | 广州品唯软件有限公司 | A kind of Method of Commodity Recommendation and relevant apparatus |
| CN111241382A (en) * | 2018-11-28 | 2020-06-05 | 北京京东尚科信息技术有限公司 | Data processing method and device, storage medium and electronic equipment |
| CN111275511A (en) * | 2018-12-05 | 2020-06-12 | 北京京东尚科信息技术有限公司 | Method, device, electronic equipment and medium for identifying competitive commodities |
| CN109933741B (en) * | 2019-02-27 | 2020-06-23 | 京东数字科技控股有限公司 | Method, device and storage medium for extracting user network behavior characteristics |
| CN111724221B (en) * | 2019-03-19 | 2024-07-19 | 北京京东尚科信息技术有限公司 | Method, system, electronic device and storage medium for determining commodity matching information |
| CN111797858A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Model training method, behavior prediction method, device, storage medium and equipment |
| CN110827112B (en) * | 2019-09-16 | 2023-02-03 | 中国平安人寿保险股份有限公司 | Deep learning commodity recommendation method and device, computer equipment and storage medium |
| CN112837106A (en) * | 2019-11-22 | 2021-05-25 | 上海哔哩哔哩科技有限公司 | Commodity recommendation method and device and computer equipment |
| CN111159555A (en) * | 2019-12-30 | 2020-05-15 | 北京每日优鲜电子商务有限公司 | Commodity recommendation method, commodity recommendation device, server and storage medium |
| CN113139767A (en) * | 2020-01-20 | 2021-07-20 | 菜鸟智能物流控股有限公司 | Logistics object allocation method and device, electronic equipment and computer-readable storage medium |
| CN111429203A (en) * | 2020-03-02 | 2020-07-17 | 北京明略软件系统有限公司 | Commodity recommendation method and device based on user behavior data |
| CN113393281A (en) * | 2020-03-11 | 2021-09-14 | 北京沃东天骏信息技术有限公司 | Method and device for processing request |
| CN113743636B (en) * | 2020-05-28 | 2024-05-14 | 北京达佳互联信息技术有限公司 | Target operation prediction method and device, electronic equipment and storage medium |
| US20210390436A1 (en) * | 2020-06-11 | 2021-12-16 | Sap Se | Determining Categories For Data Objects Based On Machine Learning |
| CN112036981B (en) * | 2020-09-02 | 2021-09-17 | 珠海必要工业科技股份有限公司 | Method, device, equipment and medium for providing target comparison commodities |
| CN113763085B (en) * | 2020-09-23 | 2024-07-19 | 京东城市(北京)数字科技有限公司 | Information pushing method, system, electronic equipment and storage medium |
| CN112149037B (en) * | 2020-09-28 | 2024-03-19 | 微梦创科网络科技（中国）有限公司 | Method and system for identifying abnormal attention in real time based on logistic regression |
| CN113298600A (en) * | 2020-09-28 | 2021-08-24 | 阿里巴巴集团控股有限公司 | Method and device for browsing commodity information |
| CN112149003B (en) * | 2020-10-28 | 2023-06-20 | 浙江集享电子商务有限公司 | Commodity community recommendation method and device and computer equipment |
| CN113761332A (en) * | 2020-11-10 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Data processing method, device, equipment and computer readable storage medium |
| CN113763012A (en) * | 2020-12-29 | 2021-12-07 | 北京京东尚科信息技术有限公司 | Model training method, similarity calculation device, similarity calculation equipment and similarity calculation medium |
| CN112767085B (en) * | 2021-01-22 | 2024-05-24 | 武汉蔚来能源有限公司 | Commodity similarity analysis and commodity recommendation method and device and computer storage medium |
| CN113781152A (en) * | 2021-02-08 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Object recommendation method and device |
| CN113254664B (en) * | 2021-05-14 | 2022-05-24 | 震坤行工业超市(上海)有限公司 | Enterprise-oriented item recommendation method and device and storage medium |
| CN113312555B (en) * | 2021-06-15 | 2023-10-03 | 北京百度网讯科技有限公司 | Conversion rate prediction method, conversion rate prediction device, electronic equipment and storage medium |
| CN113469786A (en) * | 2021-06-29 | 2021-10-01 | 深圳市点购电子商务控股股份有限公司 | Method and device for recommending articles, computer equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102156932A (en) * | 2010-02-11 | 2011-08-17 | 阿里巴巴集团控股有限公司 | Prediction method and device for secondary purchase intention of customers |
| CN103489117A (en) * | 2012-06-12 | 2014-01-01 | 深圳市腾讯计算机系统有限公司 | Method and system for information releasing |
| CN105469263A (en) * | 2014-09-24 | 2016-04-06 | 阿里巴巴集团控股有限公司 | Commodity recommendation method and device |
| CN105787770A (en) * | 2016-04-27 | 2016-07-20 | 上海遥薇(集团)有限公司 | Non-negative matrix factorization (NMF) algorithm-based big data commodity and service recommending method and system |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105528374A (en) * | 2014-10-21 | 2016-04-27 | 苏宁云商集团股份有限公司 | A commodity recommendation method in electronic commerce and a system using the same |
| CN104331816B (en) * | 2014-10-28 | 2017-10-20 | 常州大学 | Knowledge based learns big data user's purchase intention Forecasting Methodology with secret protection |
| CN104866540B (en) * | 2015-05-04 | 2018-04-27 | 华中科技大学 | A kind of personalized recommendation method based on group of subscribers behavioural analysis |
| CN105335875A (en) * | 2015-10-30 | 2016-02-17 | 小米科技有限责任公司 | Purchasing power prediction method and purchasing power prediction device |
| CN106127525A (en) * | 2016-06-27 | 2016-11-16 | 浙江大学 | A kind of TV shopping Method of Commodity Recommendation based on sorting algorithm |
-
2017
- 2017-03-29 CN CN201710200470.6A patent/CN108665329B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102156932A (en) * | 2010-02-11 | 2011-08-17 | 阿里巴巴集团控股有限公司 | Prediction method and device for secondary purchase intention of customers |
| CN103489117A (en) * | 2012-06-12 | 2014-01-01 | 深圳市腾讯计算机系统有限公司 | Method and system for information releasing |
| CN105469263A (en) * | 2014-09-24 | 2016-04-06 | 阿里巴巴集团控股有限公司 | Commodity recommendation method and device |
| CN105787770A (en) * | 2016-04-27 | 2016-07-20 | 上海遥薇(集团)有限公司 | Non-negative matrix factorization (NMF) algorithm-based big data commodity and service recommending method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108665329A (en) | 2018-10-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108665329B (en) | Commodity recommendation method based on user browsing behavior | |
| US11107118B2 (en) | Management of the display of online ad content consistent with one or more performance objectives for a webpage and/or website | |
| CN107729937B (en) | Method and device for determining user interest tag | |
| CN105989004B (en) | Information delivery preprocessing method and device | |
| JP6134444B2 (en) | Method and system for recommending information | |
| US20180047071A1 (en) | System and methods for aggregating past and predicting future product ratings | |
| CN111080398B (en) | Commodity recommendation method, commodity recommendation device, computer equipment and storage medium | |
| CN111095330B (en) | Machine learning method and system for predicting online user interactions | |
| CN112148973B (en) | Data processing method and device for information push | |
| CN113077317B (en) | Article recommendation method, device, equipment and storage medium based on user data | |
| CN112365283B (en) | Coupon issuing method and device, terminal equipment and storage medium | |
| US20180053199A1 (en) | Auto-segmentation | |
| CN112001754A (en) | User portrait generation method, device, equipment and computer readable medium | |
| JP6679451B2 (en) | Selection device, selection method, and selection program | |
| US10198762B1 (en) | Ordering search results to maximize financial gain | |
| CN112925973B (en) | Data processing method and device | |
| US20140156394A1 (en) | Targeted Advertisement Generation For Travelers | |
| CN110348947B (en) | Object recommendation method and device | |
| JP5914549B2 (en) | Information processing apparatus and information analysis method | |
| WO2020221022A1 (en) | Service object recommendation method | |
| US20210217053A1 (en) | Methods and apparatuses for selecting advertisements using semantic matching | |
| CN112598472A (en) | Product recommendation method, device, system, medium and program product | |
| CN110689402A (en) | Method and device for recommending merchants, electronic equipment and readable storage medium | |
| JP2017027359A (en) | Device, method, and program for processing information | |
| CN111680213B (en) | Information recommendation method, data processing method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |