CN110264311B - Business promotion information accurate recommendation method and system based on deep learning - Google Patents
Business promotion information accurate recommendation method and system based on deep learning Download PDFInfo
- Publication number
- CN110264311B CN110264311B CN201910461767.7A CN201910461767A CN110264311B CN 110264311 B CN110264311 B CN 110264311B CN 201910461767 A CN201910461767 A CN 201910461767A CN 110264311 B CN110264311 B CN 110264311B
- Authority
- CN
- China
- Prior art keywords
- data set
- word
- neural network
- promotion information
- lstm neural
- 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 65
- 238000013135 deep learning Methods 0.000 title claims abstract description 33
- 238000012549 training Methods 0.000 claims description 92
- 239000013598 vector Substances 0.000 claims description 87
- 238000013528 artificial neural network Methods 0.000 claims description 84
- 238000012360 testing method Methods 0.000 claims description 57
- 238000004140 cleaning Methods 0.000 claims description 21
- 230000015654 memory Effects 0.000 claims description 21
- 238000004422 calculation algorithm Methods 0.000 claims description 19
- 238000007781 pre-processing Methods 0.000 claims description 18
- 230000009471 action Effects 0.000 claims description 16
- 230000009467 reduction Effects 0.000 claims description 16
- 238000013507 mapping Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 11
- 230000011218 segmentation Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 10
- 238000010187 selection method Methods 0.000 claims description 8
- 238000002790 cross-validation Methods 0.000 claims description 6
- 238000012880 independent component analysis Methods 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 5
- 238000007418 data mining Methods 0.000 claims description 4
- 238000013480 data collection Methods 0.000 claims description 3
- 238000013501 data transformation Methods 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 3
- 238000012847 principal component analysis method Methods 0.000 claims 2
- 238000012163 sequencing technique Methods 0.000 claims 2
- 238000011161 development Methods 0.000 abstract description 4
- 239000000463 material Substances 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000011160 research Methods 0.000 description 4
- 230000018109 developmental process Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 238000011144 upstream manufacturing Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method and a system for accurately recommending business promotion information based on deep learning. The method and the system are a set of formed, efficient and regional industry-oriented recommendation system; in each area, many regional enterprises are concentrated in the same area, and the development clients are close to saturation under the line. And certain manpower and material resources cost is needed for digging potential customers offline, and the success rate is not high. Such enterprises urgently need a set of formed systems for guidance, and the cost required by developing customers is saved.
Description
Technical Field
The disclosure relates to the technical field of machine learning recommendation algorithms and deep learning, in particular to a method and a system for accurately recommending business promotion information based on deep learning.
Background
The traditional recommendation system is generally used for recommending between commodities and customers, and recommending different commodities for different customers. While the traditional recommendation algorithm generally has: content Based recommendations (CB), collaborative Filtering (CF), hybrid recommendation methods, etc. The traditional recommendation mode is abandoned, recommendation between upstream enterprises and downstream enterprises is achieved, and a deep learning-based recommendation algorithm is used for solving complex relationships between the upstream enterprises and the downstream enterprises.
The conventional recommendation system at present has the following two problems:
1) The research field of the domestic current-stage recommendation algorithm mainly focuses on accurate recommendation of commodities, and the research on customer recommendation is less. The algorithm of the domestic research recommendation system generally adopts a machine learning algorithm, but for the increasingly complex enterprise relationships and diversified data relationships, the learning efficiency of machine learning gradually cannot meet the requirements.
2) The method is separated from the traditional data mining frame, data is automatically captured and screened by the webpage, the cost of collecting data is reduced or the dilemma of lacking data is solved
Disclosure of Invention
In order to solve the problems, the invention provides a technical scheme of a deep learning-based business promotion information accurate recommendation method and system, an LSTM neural network is trained through a training sample data set, the accuracy of the LSTM neural network is tested through a testing sample data set, the trained LSTM neural network is obtained and used as a recommendation system, a classifier of the recommendation system with complete functions is obtained, based on the deep learning and network technology, the cost required by an enterprise in developing a client is reduced, the time required by developing the client is shortened, and the accuracy of finding the client is improved.
In order to achieve the above object, according to an aspect of the present disclosure, there is provided a deep learning-based business promotion information accurate recommendation method, including the steps of:
step 1, collecting commercial promotion information data;
step 2, preprocessing and cleaning the collected commercial promotion information data to obtain a commercial promotion information data set;
step 3, performing dimensionality reduction and feature selection on the commercial promotion information data set;
step 4, dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set;
step 5, obtaining a word vector for training by a training sample data set and a test sample data set through a word2vec model;
and 6, training an LSTM (Long Short-Term Memory) neural network through the training sample data set and testing the accuracy of the LSTM neural network through the test sample data set to obtain the trained LSTM neural network as a recommendation system.
Further, in step 1, the method for collecting the commercial promotion information data includes, but is not limited to: collecting an open-source data set website such as a kaggle data set as commercial promotion information data; capturing a data set from a Taobao-like shop webpage or a same-city transaction information network through a crawler by using a web crawler technology after secondary development to obtain commercial promotion information data; and acquiring required information from the webpage backup in the txt format reserved in the hectogram as business promotion information data.
Further, in step 2, the method of preprocessing and cleaning the collected commercial promotion information data to obtain the commercial promotion information data set is that, because the obtained commercial promotion information data is extremely large, mixed or even useless, preprocessing is required, and the preprocessing is necessary processing such as auditing, screening, sorting and the like before classifying or grouping the collected commercial promotion information data, namely data auditing completeness and accuracy, data screening, data sorting, namely data cleaning, data integration, data transformation and data reduction; the step of cleaning the collected commercial promotion information data refers to data cleaning, and a last program for finding and correcting recognizable errors in the data file comprises the steps of checking data consistency, processing invalid values and missing values and the like; the business promotion information data is preprocessed and cleaned, and the dirty data is converted into data meeting the data quality requirement by using mathematical statistics, data mining or predefined cleaning rules, namely a cleaned business promotion information data set.
Further, in step 3, the method for performing dimensionality reduction and feature selection on the commercial promotion information data set includes, but is not limited to, principal Component Analysis (PCA), independent Component Analysis (ICA), linear Discriminant Analysis (LDA), local Linear embedding (lop), and Linear Discriminant Analysis (lpa)Any one dimension reduction method in (Locally Linear Embedding, LLE), laplacian Eigenmaps (Laplacian Eigenmaps), multiDimensional scaling (MDS) and Equal Metric Mapping (Equal Metric Mapping); the feature selection uses an improved algorithm based on an individual optimal feature selection method, and the features are vectors obtained by text vectorization of information such as the locations, the operating ranges and the registration information of companies in the commercial promotion information data set; calculating the separability criterion value of each feature when the features are used independently by using an independent optimal feature selection algorithm, then sorting the features from large to small according to the separability criterion value, and taking the first 30 features with larger separability criterion values as feature combinations; however, in combination with the actual situation, there is a phenomenon of information missing in the collected commercial promotion information data set, so that when selecting features, it is necessary to consider a single feature information missing degree, and the improved algorithm based on the single optimal feature selection method is the following formula:
wherein, X (i) = (X (1), X (2), X (3), …, X (N)), X (i) represents the i-th feature, N is the number of features, J (X) represents the separability criterion of the feature set, N (X (i)) represents the number of data volumes of the i-th feature which are not missing, M represents the total data volume, and N (X (i))/M represents the missing degree of the i-th feature in the data, thereby improving the phenomenon of information missing.
Further, in step 4, the method for dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set includes: any one of a leave-out method, a cross-validation method and a self-service method.
The leaving method is to directly divide the commercial promotion information data set into two mutually exclusive sets, wherein one set is used as a training sample data set, and the left set is used as a test sample data set.
The cross-validation method is to divide the commercial promotion information data set into mutually exclusive subsets with equal size, namely, each subset keeps the consistency of data distribution as much as possible, namely, each subset is obtained by layered sampling, then, a union set of the subsets is used as a training sample data set each time, and the rest subset is used as a test sample data set.
The self-service method is to sample and generate a commercial promotion information data set: and selecting a sample from the commercial promotion information data set at random each time, copying the sample into a training sample data set, taking the copied sample as a test sample data set, and repeating the process for times. Wherein, some samples can appear in the data set of the commercial promotion information data set for many times as training sample data set, and the other part of samples can not appear in the data set of the commercial promotion information data set as test sample data set.
Since a class of customers with similar attributes who have purchased the particular enterprise service or product needs to be exported. The model is trained using LSTM as the core network, taking into account the different number of attributes that may be attributed to the data set. LSTM is a special type of RNN (Current Neural Network) that can learn long-term dependency information. The LSTM controls the discarding of useless information or the improvement of the proportion of beneficial information through a gate, meanwhile, a memory cell (cell) is added in the model, relevant information before memory is collected, and the function of forgetting or memorizing is realized.
Further, in step 5, the method for obtaining the word vector for training by the training sample data set and the test sample data set through the word2vec model includes the following steps:
step 5.1, word segmentation: due to the particularity of Chinese, words in the sentences in the commercial promotion information are divided through a word division library to obtain a word library, wherein the word division library comprises but is not limited to a Jieba word library, an IK word library, an mmseg word library and a word library;
step 5.2, counting word frequency: traversing the word bank formed after word segmentation in the step 5.1, counting the frequency of the appeared words and numbering the words;
and 5.3, constructing a tree result: constructing a Huffman tree according to the occurrence probability of each word in the occurrence step 5.2;
step 5.4, generating a binary code where the node is: converting the occurrence probability of each word into binary codes to represent each node in the Huffman tree in the step 5.3;
step 5.5, initializing the intermediate vectors of all non-leaf nodes and the word vectors in the leaf nodes: each node in the Huffman tree stores a vector with the length of m, but the meanings of the vectors in leaf nodes and non-leaf nodes are different, and the word vectors of each word are stored in the leaf nodes and are used as the input of a neural network; instead, the intermediate vectors are stored in the leaf nodes, correspond to the parameters of the hidden layer in the neural network, and determine the classification result together with the input;
step 5.6, training the intermediate vector and the word vector: adding word vectors Of n-1 Words near the word A by using a CBOW (Continuous Bag-Of-Words Model) Model or Skip-Gram Model to serve as system input, sequentially classifying the word vectors according to binary codes generated by the word A in step 5.4, training intermediate vectors and the word vectors according to classification results, and finally obtaining corresponding word vectors Of the commercial promotion information; the word A is a word, and the training process mainly comprises three stages of an input layer (input), a mapping layer (project) and an output layer (output); the input layer is a word vector of n-1 words around a certain word A (word A). If n takes 5, the word A (which can be denoted as w (t)), the first two and the last two words are w (t-2), w (t-1), w (t + 1), w (t + 2). Correspondingly, the word vectors for those 4 words are denoted as v (w (t-2)), v (w (t-1)), v (w (t + 1)), and v (w (t + 2)). It is relatively simple to add those n-1 word vectors from the input layer to the mapping layer.
Further, in step 6, the method for training the LSTM neural network by training the sample data set and testing the accuracy of the LSTM neural network by testing the sample data set to obtain the trained LSTM neural network as the recommendation system includes the following steps:
step 6.1, training the LSTM neural network by using the word vectors: the training sample data set passes through a forgetting gate of an LSTM neural network, information discarding action of the LSTM neural network is started, the information discarding action is realized by a sigmoid layer in the forgetting gate, previous output of the sigmoid layer and input of a current word vector are checked, whether information learned in a previous state is reserved is determined, and the LSTM neural network comprises an input gate, a forgetting gate and an output gate;
step 6.2, the training sample data set after information discarding passes through an input gate of the LSTM neural network, information updating action of the LSTM neural network is started, the information updating action is realized by a sigmoid layer in the input gate, and then each cell state of the LSTM neural network is changed by a tanh layer, and new knowledge is learned;
step 6.3, the training sample data set after the information updating passes through an output gate of the LSTM neural network, and a vector is output, wherein the vector depends on the cell state in the step 6.2; firstly, operating a sigmoid layer to obtain a vector to determine an output part of a cell state, processing the cell state through a tanh layer, and multiplying the cell state by the output (vector) of a sigmoid gate to obtain output information of an LSTM network;
and 6.4, inputting the LSTM network obtained in the step 6.3 by using the test data to obtain an output result, comparing the output result with the label in the test data to verify the accuracy of the network, finishing training to obtain a trained LSTM neural network if the accuracy meets the requirement, and taking the trained LSTM neural network as a recommendation system.
The invention also provides a deep learning-based accurate commercial promotion information recommendation system, which comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer program to run in the units of the following system:
the data collection and collection unit is used for collecting commercial promotion information data;
the data preprocessing unit is used for preprocessing and cleaning the collected commercial promotion information data to obtain a commercial promotion information data set;
the characteristic selection unit is used for performing dimensionality reduction and characteristic selection on the commercial promotion information data set;
the training sample dividing unit is used for dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set;
the vectorization unit is used for obtaining a word vector used for training through a word2vec model by a training sample data set and a test sample data set;
and the recommendation system obtaining unit is used for training the LSTM neural network through the training sample data set and testing the accuracy of the LSTM neural network through the test sample data set to obtain the trained LSTM neural network as a recommendation system.
Advantageous effects of the disclosure
The invention provides a deep learning-based accurate commercial promotion information recommendation method and system, which comprises the following steps:
1) An online service system which is not formed in the mining aspect of potential customers is aimed at, and the offline customer recommendation based on the social network basically reaches the utmost;
2) Although some enterprises research and develop systems aiming at the situations, the simple system can not adapt to various complex situations in reality and has little work effect. According to survey, the domestic system basically has no company which specially carries out information recommendation aiming at companies or enterprises, even has no set of formed, efficient and regional industry-oriented recommendation system;
3) In each region, a plurality of regional enterprises are concentrated in the same region, and the development clients are close to saturation in a line. And certain manpower and material resources cost is needed for digging potential customers off line, and the success rate is not high. The enterprises urgently need a set of formed systems for guidance, so that the cost required by developing customers is saved;
4) The LSTM has good effect on the aspect of natural language processing, has high accuracy and self-adaptive capacity when being applied to a language translator, can perfectly adapt to various complex conditions in reality, and is suitable for serving as a core network of the product.
Drawings
The foregoing and other features of the present disclosure will be more readily apparent from the detailed description of the embodiments shown in the accompanying drawings in which like reference numerals refer to the same or similar elements, and it will be apparent that the drawings in the following description are merely some examples of the disclosure, and that other drawings may be derived by those skilled in the art without inventive faculty, and wherein:
FIG. 1 is a flow chart of a deep learning-based method for accurately recommending business promotion information;
fig. 2 is a diagram illustrating a deep learning-based accurate commercial promotion information recommendation system.
Detailed Description
The conception, specific structure and technical effects of the present disclosure will be clearly and completely described below in conjunction with the embodiments and the accompanying drawings to fully understand the objects, aspects and effects of the present disclosure. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
Fig. 1 is a flowchart illustrating a method for accurately recommending deep learning-based business promotion information according to the present disclosure, and fig. 1 is combined to describe a method for accurately recommending deep learning-based business promotion information according to an embodiment of the present disclosure.
The utility model provides a commercial promotion information accurate recommendation method based on deep learning, specifically includes the following steps:
step 1, collecting commercial promotion information data;
step 2, preprocessing and cleaning the collected commercial promotion information data to obtain a commercial promotion information data set;
step 3, performing dimensionality reduction and feature selection on the commercial promotion information data set;
step 4, dividing a commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set;
step 5, obtaining a word vector for training by a training sample data set and a test sample data set through a word2vec model;
and 6, training an LSTM (Long Short-Term Memory) neural network through the training sample data set and testing the accuracy of the LSTM neural network through the test sample data set to obtain the trained LSTM neural network as a recommendation system.
Further, in step 1, the method for collecting the business promotion information data includes, but is not limited to: collecting an open-source data set website such as a kaggle data set as commercial promotion information data; capturing a data set from a Taobao-like shop webpage or a same-city transaction information network through a crawler by using a web crawler technology after secondary development to obtain commercial promotion information data; and acquiring required information from the webpage backup in the txt format reserved in the hectogram as business promotion information data.
Further, in step 2, the method of preprocessing and cleaning the collected commercial promotion information data to obtain the commercial promotion information data set is that, because the obtained commercial promotion information data is extremely large, mixed or even useless, preprocessing is required, and the preprocessing is necessary processing such as auditing, screening, sorting and the like before classifying or grouping the collected commercial promotion information data, namely data auditing completeness and accuracy, data screening, data sorting, namely data cleaning, data integration, data transformation and data reduction; the step of cleaning the collected commercial promotion information data refers to data cleaning, and a last program for finding and correcting recognizable errors in the data file comprises the steps of checking data consistency, processing invalid values and missing values and the like; the business promotion information data is preprocessed and cleaned, and the dirty data is converted into data meeting the data quality requirement by using mathematical statistics, data mining or predefined cleaning rules, namely a cleaned business promotion information data set.
Further, in step 3, the method for performing dimensionality reduction and feature selection on the commercial promotion information data set includes, but is not limited to, any one of a Principal Component Analysis (PCA), independent Component Analysis (ICA), linear Discriminant Analysis (LDA), local Linear Embedding (LLE), laplacian feature mapping, multidimensional scaling (MDS), and isometry mapping; and the feature selection uses an improved algorithm based on a single optimal feature selection method, wherein the features are information such as the locations, the operating ranges, the registration information and the like of the companies in the commercial promotion information data setVectorizing the text; calculating the separability criterion value of each feature when the feature is used independently by using an independent optimal feature selection algorithm, then sorting the separability criterion values from large to small, and taking the first 30 features with larger separability criterion values as feature combinations; however, in combination with practical situations, there is a phenomenon of information missing in the collected commercial promotion information data set, so that when selecting features, a single feature information missing degree needs to be considered, and the improved algorithm based on the single optimal feature selection method is the following formula:
wherein, X (i) = (X (1), X (2), X (3), …, X (N)), X (i) represents the ith feature, N is the number of features, J (X) represents the separability criterion of the feature set, N (X (i)) represents the number of data volumes of the ith feature which are not missing, M represents the total data volume, and N (X (i))/M represents the missing degree of the ith feature in the data.
Further, in step 4, the method for dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set includes: any one of a leave-out method, a cross-validation method and a self-service method.
The method for setting out the business promotion information data sets comprises the steps of directly dividing the business promotion information data sets into two mutually exclusive sets, wherein one set serves as a training sample data set, and the set which is set out serves as a test sample data set.
The cross-validation method is that the commercial promotion information data set is divided into mutually exclusive subsets with equal size, namely, each subset keeps the consistency of data distribution as much as possible, namely, each subset is obtained by layered sampling, then, the union set of the subsets is used as a training sample data set every time, and the rest subset is used as a test sample data set.
The self-service method is to sample and generate a commercial promotion information data set: and selecting a sample from the commercial promotion information data set at random each time, copying the sample into a training sample data set, taking the copied sample as a test sample data set, and repeating the process for times. Wherein, some samples can appear in the data set of the commercial promotion information data set for many times as training sample data set, and the other part of samples can not appear in the data set of the commercial promotion information data set as test sample data set.
Since a class of customers with similar attributes who have purchased the particular business service or product needs to be exported. The model is trained using the LSTM network as the core network, taking into account the different number of attributes that may be attributed to the data set. LSTM is a special type of RNN that can learn long-term dependency information. The LSTM controls the discarding of useless information or the improvement of the proportion of beneficial information through a gate, meanwhile, memory cells are added in the model, relevant information before memory is collected, and the function of forgetting or memorizing is realized.
Further, in step 5, the method for obtaining the word vector for training by the training sample data set and the test sample data set through the word2vec model includes the following steps:
step 5.1, word segmentation: due to the particularity of Chinese, words in the sentences in the commercial promotion information are divided through a word division library to obtain a word library, wherein the word division library comprises but is not limited to a Jieba word library, an IK word library, an mmseg word library and a word library;
step 5.2, counting word frequency: traversing the word bank formed after word segmentation in the step 5.1, counting the frequency of the appeared words and numbering the words;
step 5.3, constructing a tree result: constructing a Huffman tree according to the occurrence probability of each word in the occurrence step 5.2;
step 5.4, generating a binary code where the node is: converting the occurrence probability of each word into binary codes to represent each node in the Huffman tree in the step 5.3;
step 5.5, initializing the intermediate vectors of all non-leaf nodes and the word vectors in the leaf nodes: each node in the Huffman tree stores a vector with the length of m, but the meanings of the vectors in leaf nodes and non-leaf nodes are different, and the word vectors of each word are stored in the leaf nodes and are used as the input of a neural network; instead, the intermediate vectors are stored in the leaf nodes, correspond to the parameters of the hidden layer in the neural network, and determine the classification result together with the input;
step 5.6, training the intermediate vector and the word vector: adding word vectors Of n-1 Words near the word A by using a CBOW (Continuous Bag-Of-Words Model) Model or Skip-Gram Model to serve as system input, sequentially classifying the word vectors according to binary codes generated by the word A in step 5.4, training intermediate vectors and the word vectors according to classification results, and finally obtaining corresponding word vectors Of the commercial promotion information; the word A is a word, and the training process mainly comprises three stages of an input layer (input), a mapping layer (project) and an output layer (output); the input layer is a word vector of n-1 words around a certain word A (word A). If n takes 5, the word A (which can be denoted as w (t)), the first two and the last two words are w (t-2), w (t-1), w (t + 1), w (t + 2). Correspondingly, the word vectors for those 4 words are denoted as v (w (t-2)), v (w (t-1)), v (w (t + 1)), and v (w (t + 2)). It is relatively simple to go from the input layer to the mapping layer by adding those n-1 word vectors.
Further, in step 6, the method for training the LSTM neural network by training the sample data set and testing the accuracy of the LSTM neural network by testing the sample data set to obtain the trained LSTM neural network as the recommendation system includes the following steps:
step 6.1, training the LSTM neural network by using the word vectors: the training sample data set passes through a forgetting gate of an LSTM neural network, information discarding action of the LSTM neural network is started, the information discarding action is realized by a sigmoid layer in the forgetting gate, previous output of the sigmoid layer and input of a current word vector are checked, whether information learned in a previous state is reserved is determined, and the LSTM neural network comprises an input gate, a forgetting gate and an output gate;
step 6.2, the training sample data set after information discarding passes through an input gate of the LSTM neural network, information updating action of the LSTM neural network is started, the information updating action is realized by a sigmoid layer in the input gate, and then each cell state of the LSTM neural network is changed by a tanh layer, and new knowledge is learned;
step 6.3, the training sample data set after the information updating passes through an output gate of the LSTM neural network, and a vector is output, wherein the vector depends on the cell state in the step 6.2; firstly, operating a sigmoid layer to obtain a vector to determine an output part of a cell state, processing the cell state through a tanh layer, and multiplying the cell state by the output (vector) of a sigmoid gate to obtain output information of an LSTM network;
and 6.4, inputting the LSTM network obtained in the step 6.3 by using the test data to obtain an output result, comparing the output result with the label in the test data to verify the accuracy of the network, finishing training to obtain a trained LSTM neural network if the accuracy meets the requirement, and taking the trained LSTM neural network as a recommendation system.
By means of the recommendation system, how the method is used for accurately recommending the clients to the enterprise can be demonstrated, and the enterprise is helped to develop a large number of potential and accurate clients more efficiently. By using the thinking of 'internet + finance' and based on deep learning and network technology, the cost required by enterprises in developing clients is reduced, the time required by developing clients is shortened, and the accuracy of finding clients is improved.
The accuracy rate (0.12) of the mined data of the accurate commercial promotion information recommendation system based on deep learning is far higher than the accuracy rate (0.067) of the traditional content-based method, and the retention rate is also obviously improved; and hot commercial promotion information can be found more easily. The LSTM deep neural network is applied to commercial popularization information mining, irrelevant noise information is forgotten, the memory of strong relevant information is deepened, and the LSTM deep neural network has the advantages and disadvantages in algorithm and has the preferred selection characteristic. The Huffman tree is adopted to perform related business promotion information word segmentation, the word segmentation speed is greatly improved, and the word segmentation calculation time is about 1/20 of the general exhaustive expression.
An embodiment of the present disclosure provides an accurate recommendation system of business promotion information based on deep learning, as shown in fig. 2, is an accurate recommendation system diagram of business promotion information based on deep learning of the present disclosure, and an accurate recommendation system of business promotion information based on deep learning of the embodiment includes: the system comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to realize the steps in the embodiment of the deep learning based commercial promotion information accurate recommendation system.
The system comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer program to run in the units of the system:
the data collection and collection unit is used for collecting commercial promotion information data;
the data preprocessing unit is used for preprocessing and cleaning the collected commercial promotion information data to obtain a commercial promotion information data set;
the characteristic selection unit is used for performing dimensionality reduction and characteristic selection on the commercial promotion information data set;
the training sample dividing unit is used for dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set;
the vectorization unit is used for obtaining a word vector used for training through a word2vec model by a training sample data set and a test sample data set;
and the recommendation system obtaining unit is used for training the LSTM neural network through the training sample data set and testing the accuracy of the LSTM neural network through the test sample data set to obtain the trained LSTM neural network as a recommendation system.
The deep learning-based accurate commercial promotion information recommendation system can be operated in computing equipment such as desktop computers, notebooks, palm computers and cloud servers. The business promotion information accurate recommendation system based on deep learning can be operated by a system comprising but not limited to a processor and a memory. It will be understood by those skilled in the art that the example is only an example of a deep learning-based accurate commercial promotion information recommendation system, and does not constitute a limitation of a deep learning-based accurate commercial promotion information recommendation system, and may include more or less components than a certain proportion, or some components in combination, or different components, for example, the deep learning-based accurate commercial promotion information recommendation system may further include an input-output device, a network access device, a bus, etc.
The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can be any conventional processor, and the processor is a control center of the deep learning-based accurate recommendation system for business promotion information, and various interfaces and lines are used for connecting all parts of the whole system capable of operating the deep learning-based accurate recommendation system for business promotion information.
The memory can be used for storing the computer program and/or the module, and the processor can realize various functions of the deep learning-based accurate commercial promotion information recommendation system by running or executing the computer program and/or the module stored in the memory and calling the data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, a phonebook, etc.) created according to the use of the cellular phone, and the like. In addition, the memory may include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), at least one magnetic disk storage device, a Flash memory device, or other volatile solid state storage device.
While the present disclosure has been described in considerable detail and with particular reference to a few illustrative embodiments thereof, it is not intended to be limited to any such details or embodiments or any particular embodiments, but it is to be construed as effectively covering the intended scope of the disclosure by providing a broad, potential interpretation of such claims in view of the prior art with reference to the appended claims. Furthermore, the foregoing describes the disclosure in terms of embodiments foreseen by the inventor for which an enabling description was available, notwithstanding that insubstantial modifications of the disclosure, not presently foreseen, may nonetheless represent equivalent modifications thereto.
Claims (3)
1. A deep learning-based accurate business promotion information recommendation method is characterized by comprising the following steps:
step 1, collecting commercial promotion information data;
step 2, preprocessing and cleaning the collected commercial promotion information data to obtain a commercial promotion information data set;
step 3, performing dimensionality reduction and feature selection on the commercial promotion information data set;
step 4, dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set;
step 5, obtaining a word vector for training by a training sample data set and a test sample data set through a word2vec model;
in step 5, the method for obtaining the word vector for training by the training sample data set and the test sample data set through the word2vec model comprises the following steps:
step 5.1, word segmentation: due to the particularity of Chinese, words in the sentences in the commercial promotion information are divided through a word division library to obtain a word library, wherein the word division library comprises a Jieba word library, an IK word library, an mmseg word library and a word library;
step 5.2, counting word frequency: traversing the word bank formed after word segmentation in the step 5.1, counting the frequency of the appeared words and numbering the words;
and 5.3, constructing a tree result: constructing a Huffman tree according to the occurrence probability of each word in the occurrence step 5.2;
step 5.4, generating a binary code where the node is: converting the occurrence probability of each word into binary codes to represent each node in the Huffman tree in the step 5.3;
step 5.5, initializing the intermediate vectors of all non-leaf nodes and the word vectors in the leaf nodes: each node in the Huffman tree stores a length ofmThe meaning of the vectors in the leaf nodes and the non-leaf nodes is different, and the word vectors of all words are stored in the leaf nodes and are used as the input of the neural network; instead, the intermediate vectors are stored in the leaf nodes, correspond to the parameters of the hidden layer in the neural network, and determine the classification result together with the input;
step 5.6, training the intermediate vector and the word vector: training intermediate vectors and word vectors by using a CBOW model or a Skip-Gram model, and finally obtaining corresponding word vectors of the commercial promotion information;
step 6, training the LSTM neural network through the training sample data set and testing the accuracy of the LSTM neural network through the test sample data set to obtain the trained LSTM neural network as a recommendation system;
in step 3, the method for performing dimensionality reduction and feature selection on the commercial promotion information data set comprises any one dimensionality reduction method of a principal component analysis method, independent component analysis, linear discriminant analysis, local linear embedding, laplace feature mapping, multidimensional scaling and equal metric mapping; the feature selection uses an improved algorithm based on an individual optimal feature selection method, and the features are vectors of the places, the operating ranges and the registration information of the companies in the commercial promotion information data set after text vectorization; calculating the separability criterion value of each feature when the feature is used independently by using an independent optimal feature selection algorithm, then sorting the separability criterion values from large to small, and taking the first 30 features with larger separability criterion values as feature combinations; the improved algorithm based on the single optimal feature selection method is as follows:
wherein,x(i)=（x(1),x(2),x(3),…,x(n)），x(i) Represents the firstiThe characteristics of the device are as follows,nthe number of the characteristics is the same as the number of the characteristics,J(X) A separability criterion representing the set of features,N(x(i) Is expressed asiA characteristic is not the number of missing data volumes,Mwhich represents the total amount of the data amount,N(x(i))/Mshow thatiThe degree of absence of individual features in the data;
in step 4, the method for dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set includes: any one of a leave-out method, a cross-validation method and a self-service method;
in step 6, the method for training the LSTM neural network through the training sample data set and testing the accuracy of the LSTM neural network through the testing sample data set to obtain the trained LSTM neural network as the recommendation system comprises the following steps:
step 6.1, training the LSTM neural network by using the word vectors: the training sample data set passes through a forgetting gate of an LSTM neural network, information discarding action of the LSTM neural network is started, the information discarding action is realized by a sigmoid layer in the forgetting gate, previous output of the sigmoid layer and input of a current word vector are checked, whether information learned in a previous state is reserved is determined, and the LSTM neural network comprises an input gate, a forgetting gate and an output gate;
step 6.2, the training sample data set after information discarding passes through an input gate of the LSTM neural network, information updating action of the LSTM neural network is started, the information updating action is realized by a sigmoid layer in the input gate, and then each cell state of the LSTM neural network is changed by a tanh layer, and new knowledge is learned;
step 6.3, the training sample data set after the information updating passes through an output gate of the LSTM neural network, and a vector is output, wherein the vector depends on the cell state in the step 6.2; firstly, operating a sigmoid layer to obtain a vector to determine an output part of a cell state, processing the cell state through a tanh layer, and multiplying the cell state by the output of a sigmoid gate to obtain output information of an LSTM network;
and 6.4, inputting the LSTM network obtained in the step 6.3 by using the test data to obtain an output result, comparing the output result with the label in the test data to verify the accuracy of the network, finishing training to obtain a trained LSTM neural network if the accuracy meets the requirement, and taking the trained LSTM neural network as a recommendation system.
2. The method for accurately recommending business promotion information based on deep learning according to claim 1, wherein in step 2, the method for preprocessing and cleaning the collected business promotion information data to obtain the business promotion information data set comprises preprocessing, wherein the preprocessing is the processing of auditing, screening and sequencing before classifying or grouping the collected business promotion information data, namely data auditing and accuracy, data screening and data sequencing, namely data cleaning, data integration, data transformation and data reduction; the step of cleaning the collected commercial promotion information data refers to data cleaning, and a last program for finding and correcting recognizable errors in the data file comprises the steps of checking data consistency and processing invalid values and missing values; the business promotion information data is preprocessed and cleaned, and the dirty data is converted into data meeting the data quality requirement by using mathematical statistics, data mining or predefined cleaning rules, namely a cleaned business promotion information data set.
3. A system for accurately recommending business promotion information based on deep learning, the system comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the computer program to run in the units of the following system:
the data collection and collection unit is used for collecting commercial promotion information data;
the data preprocessing unit is used for preprocessing and cleaning the collected commercial promotion information data to obtain a commercial promotion information data set;
the characteristic selection unit is used for performing dimensionality reduction and characteristic selection on the commercial promotion information data set;
the training sample dividing unit is used for dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set;
the vectorization unit is used for obtaining a word vector used for training through a word2vec model by a training sample data set and a test sample data set;
the method for obtaining the word vector for training by the training sample data set and the test sample data set through the word2vec model comprises the following steps:
step 5.1, word segmentation: due to the particularity of Chinese, words in the sentences in the commercial promotion information are divided through a word division library to obtain a word library, wherein the word division library comprises a Jieba word library, an IK word library, an mmseg word library and a word library;
step 5.2, word frequency statistics: traversing the word bank formed after word segmentation in the step 5.1, counting the frequency of the appeared words and numbering the words;
and 5.3, constructing a tree result: constructing a Huffman tree according to the occurrence probability of each word in the occurrence step 5.2;
step 5.4, generating a binary code where the node is: converting the occurrence probability of each word into binary codes to represent each node in the Huffman tree in the step 5.3;
step 5.5, initializing the intermediate vectors of all non-leaf nodes and the word vectors in the leaf nodes: each node in the Huffman tree stores a length ofmThe vectors in the leaf nodes and the non-leaf nodes have different meanings, and the word vectors of all words are stored in the leaf nodes and are used as the input of the neural network; instead, the intermediate vectors stored in the leaf nodes correspond to the parameters of the hidden layer in the neural network and determine the classification result together with the input;
step 5.6, training the intermediate vector and the word vector: training intermediate vectors and word vectors by using a CBOW model or a Skip-Gram model, and finally obtaining corresponding word vectors of the commercial promotion information;
the recommendation system obtaining unit is used for training the LSTM neural network through the training sample data set and testing the accuracy of the LSTM neural network through the test sample data set to obtain the trained LSTM neural network as a recommendation system;
the method for performing dimensionality reduction and feature selection on the commercial promotion information data set comprises any dimensionality reduction method of a principal component analysis method, an independent component analysis method, a linear discriminant analysis method, local linear embedding, laplace feature mapping, multidimensional scaling and equal-metric mapping; the feature selection uses an improved algorithm based on an individual optimal feature selection method, and the features are vectors obtained by text vectorization of information such as the locations, the operating ranges and the registration information of companies in the commercial promotion information data set; calculating the separability criterion value of each feature when the feature is used independently by using an independent optimal feature selection algorithm, then sorting the separability criterion values from large to small, and taking the first 30 features with larger separability criterion values as feature combinations; however, in combination with practical situations, there is a phenomenon of information missing in the collected commercial promotion information data set, so that when selecting features, a single feature information missing degree needs to be considered, and the improved algorithm based on the single optimal feature selection method is the following formula:
wherein,x(i)=（x(1),x(2),x(3),…,x(n)），x(i) Represents the firstiThe characteristics of the device are as follows,nthe number of the characteristics is the same as the number of the characteristics,J(X) A separability criterion representing the set of features,N(x(i) Is expressed asiA characteristic is not the number of missing data volumes,Mwhich represents the total amount of data volume,N(x(i))/Mshow thatiThe degree of absence of individual features in the data;
the method for dividing the commercial promotion information data set obtained by feature selection into a training sample data set and a test sample data set comprises the following steps: any one of a leave-out method, a cross-validation method and a self-service method;
the method for training the LSTM neural network through the training sample data set and testing the accuracy of the LSTM neural network through the testing sample data set to obtain the trained LSTM neural network as the recommendation system comprises the following steps:
step 6.1, training the LSTM neural network by using the word vectors: the training sample data set passes through a forgetting gate of an LSTM neural network, information discarding action of the LSTM neural network is started, the information discarding action is realized by a sigmoid layer in the forgetting gate, previous output of the sigmoid layer and input of a current word vector are checked, whether information learned in a previous state is reserved is determined, and the LSTM neural network comprises an input gate, a forgetting gate and an output gate;
step 6.2, the training sample data set after information discarding passes through an input gate of the LSTM neural network, information updating action of the LSTM neural network is started, the information updating action is realized by a sigmoid layer in the input gate, and then each cell state of the LSTM neural network is changed by a tanh layer, and new knowledge is learned;
step 6.3, the training sample data set after the information updating passes through an output gate of the LSTM neural network, and a vector is output, wherein the vector depends on the cell state in the step 6.2; firstly, operating a sigmoid layer to obtain a vector to determine an output part of a cell state, processing the cell state through a tanh layer, and multiplying the cell state by the output of a sigmoid gate to obtain output information of an LSTM network;
and 6.4, inputting the LSTM network obtained in the step 6.3 by using the test data to obtain an output result, comparing the output result with the label in the test data to verify the accuracy of the network, finishing training to obtain a trained LSTM neural network if the accuracy meets the requirement, and taking the trained LSTM neural network as a recommendation system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910461767.7A CN110264311B (en) | 2019-05-30 | 2019-05-30 | Business promotion information accurate recommendation method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910461767.7A CN110264311B (en) | 2019-05-30 | 2019-05-30 | Business promotion information accurate recommendation method and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110264311A CN110264311A (en) | 2019-09-20 |
| CN110264311B true CN110264311B (en) | 2023-04-18 |
Family
ID=67915965
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910461767.7A Active CN110264311B (en) | 2019-05-30 | 2019-05-30 | Business promotion information accurate recommendation method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110264311B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111860981B (en) * | 2020-07-03 | 2024-01-19 | 航天信息(山东)科技有限公司 | Enterprise national industry category prediction method and system based on LSTM deep learning |
| CN112465389A (en) * | 2020-12-12 | 2021-03-09 | 广东电力信息科技有限公司 | Word frequency-based similar provider recommendation method and device |
| CN113806626A (en) * | 2021-01-25 | 2021-12-17 | 北京沃东天骏信息技术有限公司 | Method and system for sending push message |
| CN118379116B (en) * | 2024-06-24 | 2024-10-22 | 南京信息工程大学 | Deep learning-based Internet data deep mining method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105869024A (en) * | 2016-04-20 | 2016-08-17 | 北京小米移动软件有限公司 | Commodity recommending method and device |
| CN107291693A (en) * | 2017-06-15 | 2017-10-24 | 广州赫炎大数据科技有限公司 | A kind of semantic computation method for improving term vector model |
| CN108427670A (en) * | 2018-04-08 | 2018-08-21 | 重庆邮电大学 | A kind of sentiment analysis method based on context word vector sum deep learning |
| CN109635204A (en) * | 2018-12-21 | 2019-04-16 | 上海交通大学 | Online recommender system based on collaborative filtering and length memory network |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9626682B2 (en) * | 2012-02-08 | 2017-04-18 | Ebay Inc. | Systems and methods for reseller discovery and analysis |
| CN107153642A (en) * | 2017-05-16 | 2017-09-12 | 华北电力大学 | A kind of analysis method based on neural network recognization text comments Sentiment orientation |
| CN108256052B (en) * | 2018-01-15 | 2023-07-11 | 成都达拓智通科技有限公司 | Tri-tracking-based potential customer identification method for automobile industry |
-
2019
- 2019-05-30 CN CN201910461767.7A patent/CN110264311B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105869024A (en) * | 2016-04-20 | 2016-08-17 | 北京小米移动软件有限公司 | Commodity recommending method and device |
| CN107291693A (en) * | 2017-06-15 | 2017-10-24 | 广州赫炎大数据科技有限公司 | A kind of semantic computation method for improving term vector model |
| CN108427670A (en) * | 2018-04-08 | 2018-08-21 | 重庆邮电大学 | A kind of sentiment analysis method based on context word vector sum deep learning |
| CN109635204A (en) * | 2018-12-21 | 2019-04-16 | 上海交通大学 | Online recommender system based on collaborative filtering and length memory network |
Non-Patent Citations (1)
| Title |
|---|
| 数据预处理;李爱国、库向阳;《数据挖掘原理、算法及应用》;20120131;第25-27页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110264311A (en) | 2019-09-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112785397B (en) | Product recommendation method, device and storage medium | |
| US11741361B2 (en) | Machine learning-based network model building method and apparatus | |
| CN110264311B (en) | Business promotion information accurate recommendation method and system based on deep learning | |
| CN110069709B (en) | Intention recognition method, device, computer readable medium and electronic equipment | |
| US20230102337A1 (en) | Method and apparatus for training recommendation model, computer device, and storage medium | |
| CN106447066A (en) | Big data feature extraction method and device | |
| US11373117B1 (en) | Artificial intelligence service for scalable classification using features of unlabeled data and class descriptors | |
| CN109582956A (en) | text representation method and device applied to sentence embedding | |
| Manovich | Can we think without categories? | |
| CN108021651B (en) | Network public opinion risk assessment method and device | |
| CN110795568A (en) | Risk assessment method and device based on user information knowledge graph and electronic equipment | |
| CN109690581B (en) | User guidance system and method | |
| CN111859149A (en) | Information recommendation method and device, electronic equipment and storage medium | |
| CN108549909B (en) | Object classification method and object classification system based on crowdsourcing | |
| US20230004979A1 (en) | Abnormal behavior detection method and apparatus, electronic device, and computer-readable storage medium | |
| CN112734569A (en) | Stock risk prediction method and system based on user portrait and knowledge graph | |
| CN114416998A (en) | Text label identification method and device, electronic equipment and storage medium | |
| CN112418320A (en) | Enterprise association relation identification method and device and storage medium | |
| CN113327132A (en) | Multimedia recommendation method, device, equipment and storage medium | |
| US11907508B1 (en) | Content analytics as part of content creation | |
| Miao et al. | Informative core identification in complex networks | |
| CN107315807B (en) | Talent recommendation method and device | |
| CN112463964B (en) | Text classification and model training method, device, equipment and storage medium | |
| CN116739649A (en) | User response potential evaluation method and device | |
| CN111767710B (en) | Indonesia emotion classification method, device, equipment and medium |
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 | ||
| CP03 | Change of name, title or address |
Address after: No.33 Guangyun Road, Shishan town, Nanhai District, Foshan City, Guangdong Province Patentee after: Foshan University Country or region after: China Address before: No.33 Guangyun Road, Shishan town, Nanhai District, Foshan City, Guangdong Province Patentee before: FOSHAN University Country or region before: China |
|
| CP03 | Change of name, title or address |