CN116739695A

CN116739695A - Electronic commerce management system and method based on big data

Info

Publication number: CN116739695A
Application number: CN202310525272.2A
Authority: CN
Inventors: 洪燕仪
Original assignee: Guangdong Bailu Electronic Technology Co ltd
Current assignee: Guangdong Bailu Electronic Technology Co ltd
Priority date: 2023-05-11
Filing date: 2023-05-11
Publication date: 2023-09-12

Abstract

The application relates to the field of big data, and particularly discloses an electronic commerce management system and method based on big data. In this way, the shopping preference of the user is accurately predicted by utilizing big data in the electronic commerce system, so that more personalized and intelligent commodity recommendation service is provided for the user.

Description

Electronic commerce management system and method based on big data

Technical Field

The present application relates to the field of big data, and more particularly, to an electronic commerce management system based on big data and a method thereof.

Background

Electronic commerce is a commercial model for commodity or service transaction by using the internet technology, and has the advantages of wide transaction range, high transaction speed, low transaction cost and the like. With the development of the internet, massive data including user data, commodity data, transaction data and the like are generated on an electronic commerce platform. These data reflect not only the consumer behavior and preferences of the user, but also the characteristics and value of the merchandise, as well as the needs and changes in the marketplace. Therefore, how to efficiently collect, store, analyze, and utilize such data is critical to improving the efficiency and level of intelligence of electronic commerce management systems.

Disclosure of Invention

The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides an electronic commerce management system and a method thereof based on big data, which comprehensively utilize historical transaction records, search historical data and browsing historical data to realize the aim of consumer shopping preference prediction by adopting an artificial intelligence technology based on deep learning. In this way, the shopping preference of the user is accurately predicted by utilizing big data in the electronic commerce system, so that more personalized and intelligent commodity recommendation service is provided for the user.

According to an aspect of the present application, there is provided an electronic commerce management system based on big data, including:

the system comprises a user data acquisition module, a user data analysis module and a user data analysis module, wherein the user data acquisition module is used for extracting relevant data of an analyzed user from an electronic commerce system, and the relevant data comprises a historical transaction record, search historical data and browsing historical data;

a user portrayal construction module for constructing a user portrayal of the analyzed user based on the related data of the analyzed user, wherein the user portrayal comprises basic properties, purchasing behavior, interests and purchasing intention;

The word embedding module is used for enabling each attribute item in the user portrait of the analyzed user to pass through a word embedding layer to obtain a plurality of attribute item word embedding vectors;

the attribute value adding module is used for respectively adding the attribute value of each attribute term in the user portrait of the analyzed user to the tail part of each attribute term word embedding vector so as to obtain a plurality of user portrait attribute word embedding vectors;

the first semantic understanding module is used for enabling the plurality of user portrait attribute words to be embedded into vectors and enabling the user portrait attribute words to pass through a context encoder based on a converter so as to obtain first-scale user portrait semantic understanding feature vectors;

the second semantic understanding module is used for arranging the plurality of user portrait attribute word embedded vectors into a two-dimensional feature matrix and then obtaining second-scale user portrait semantic understanding feature vectors through a text convolutional neural network model;

the feature fusion module is used for fusing the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector to obtain a user portrait semantic understanding feature vector; and

and the label dividing module is used for enabling the user portrait semantic understanding feature vector to pass through a classifier to obtain a classification result, and the classification result is used for representing shopping preference type labels.

In the above-mentioned sterilizing device for intensive care and nursing of I CU, the dynamic and static data association module is configured to: cascading the power time sequence input vector and the power change time sequence input vector by using the following cascading formula to obtain a power time sequence multi-scale input vector; wherein, the formula is:

V _c ＝Concat[V ₁ ，V ₂ ]

wherein V is ₁ Representing the power timing input vector, V ₂ Representing the power change timing input vector, concat [. Cndot.,)]Representing a cascade function, V _c Representing the power-sequential multiscale input vector.

In the electronic commerce management system based on big data, the word embedding module comprises: a one-dimensional unfolding unit for unfolding each attribute item in the user portrait of the analyzed user into a one-dimensional pixel vector to obtain a sequence of a plurality of one-dimensional pixel vectors; and an embedding encoding unit, configured to perform embedding encoding on one-dimensional pixel vectors in the sequence of the plurality of one-dimensional pixel vectors by using the weight matrix of the embedding layer to obtain the plurality of attribute term word embedding vectors.

In the electronic commerce management system based on big data, the first semantic understanding module includes: the query vector construction unit is used for carrying out one-dimensional arrangement on the plurality of user portrait attribute word embedded vectors so as to obtain global user portrait feature vectors; a self-attention unit, configured to calculate a product between the global user portrait feature vector and a transpose vector of each user portrait attribute word embedding vector in the plurality of user portrait attribute word embedding vectors to obtain a plurality of self-attention correlation matrices; the normalization unit is used for respectively performing normalization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices; the attention calculating unit is used for obtaining a plurality of probability values through a Softmax classification function by each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; the attention applying unit is used for weighting each user portrait attribute word embedded vector in the plurality of user portrait attribute word embedded vectors by taking each probability value in the plurality of probability values as a weight so as to obtain the plurality of context semantic user portrait feature vectors; and the cascading unit is used for cascading the plurality of context semantic user portrait feature vectors to obtain the first-scale user portrait semantic understanding feature vector.

In the electronic commerce management system based on big data, the second semantic understanding module includes: each layer of the text convolutional neural network model is used for respectively carrying out forward transfer on input data in the layers: carrying out convolution processing on input data to obtain a convolution characteristic diagram; pooling the convolution feature images based on a feature matrix to obtain pooled feature images; performing nonlinear activation on the pooled feature map to obtain an activated feature map; the output of the last layer of the text convolutional neural network is the semantic understanding feature vector of the second-scale user portrait, and the input of the first layer of the text convolutional neural network is a two-dimensional feature matrix obtained by arranging the attribute word embedded vectors of the plurality of user portrait.

In the electronic commerce management system based on big data, the feature fusion module is further configured to: carrying out deep space packaging semantic matching fusion on the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector by using the following formula to obtain the user portrait semantic understanding feature vector; wherein, the formula is:

Wherein V is ₁ Is the first scale user portrait semantic understanding feature vector, V ₂ Is the second scale user portrait semantic understanding feature vector, I.I ₁ And|| | ₂ Represents the first norm and the second norm of the vector respectively, ω and ε are the weight and bias hyper-parameters respectively, D (V ₁ ,V ₂ ) Representing a per-position distance matrix between the first scale user portrait semantic understanding feature vector and the second scale user portrait semantic understanding feature vector, and I is an identity matrix,as indicated above, V is the sum by position, the subtraction by position and the multiplication by position _c Is the user portrait semantic understanding feature vector.

In the electronic commerce management system based on big data, the label dividing module includes: the full-connection coding unit is used for carrying out full-connection coding on the user portrait semantic understanding feature vector by using a plurality of full-connection layers of the classifier so as to obtain a coded user portrait semantic understanding feature vector; and the classification result generating unit is used for enabling the encoded user portrait semantic understanding feature vector to pass through a Softmax classification function of the classifier to obtain the classification result.

According to another aspect of the present application, there is provided an electronic commerce management method based on big data, including:

Extracting relevant data of an analyzed user from an electronic commerce system, wherein the relevant data comprises historical transaction records, search historical data and browsing historical data;

constructing a user representation of the analyzed user based on the related data of the analyzed user, wherein the user representation comprises basic attributes, purchasing behavior, interests and purchasing intent;

passing each attribute term in the user portrait of the analyzed user through a term embedding layer to obtain a plurality of attribute term embedding vectors;

respectively adding the attribute values of all attribute terms in the user portrait of the analyzed user to the tail parts of all attribute term embedding vectors to obtain a plurality of user portrait attribute term embedding vectors;

embedding the plurality of user portrayal attribute words into vectors through a context encoder based on a converter to obtain first-scale user portrayal semantic understanding feature vectors;

arranging the embedding vectors of the user portrait attribute words into a two-dimensional feature matrix, and then obtaining a second-scale user portrait semantic understanding feature vector through a text convolutional neural network model;

fusing the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector to obtain a user portrait semantic understanding feature vector; and

And passing the user portrait semantic understanding feature vector through a classifier to obtain a classification result, wherein the classification result is used for representing shopping preference type labels.

According to still another aspect of the present application, there is provided an electronic apparatus including: a processor; and a memory in which computer program instructions are stored which, when executed by the processor, cause the processor to perform the big data based e-commerce management method as described above.

According to yet another aspect of the present application, there is provided a computer readable medium having stored thereon computer program instructions which, when executed by a processor, cause the processor to perform the big data based e-commerce management method as described above.

Compared with the prior art, the electronic commerce management system and the method based on big data provided by the application realize the purpose of consumer shopping preference prediction by comprehensively utilizing the historical transaction record, the search historical data and the browsing historical data by adopting the artificial intelligence technology based on deep learning. In this way, the shopping preference of the user is accurately predicted by utilizing big data in the electronic commerce system, so that more personalized and intelligent commodity recommendation service is provided for the user.

Drawings

The above and other objects, features and advantages of the present application will become more apparent by describing embodiments of the present application in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.

FIG. 1 is a block diagram of a big data based e-commerce management system in accordance with an embodiment of the present application;

FIG. 2 is a system architecture diagram of a big data based e-commerce management system in accordance with an embodiment of the present application;

FIG. 3 is a block diagram of a first semantic understanding module in a big data based e-commerce management system according to an embodiment of the present application;

FIG. 4 is a flow chart of text convolutional neural network coding in a big data based e-commerce management system in accordance with an embodiment of the present application;

FIG. 5 is a flow chart of a big data based e-commerce management method according to an embodiment of the present application;

fig. 6 is a block diagram of an electronic device according to an embodiment of the application.

Detailed Description

Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.

Summary of the application

Aiming at the technical problems, the technical conception of the application is to combine deep learning and artificial intelligence technology and comprehensively utilize historical transaction records, search historical data and browsing historical data to realize the aim of consumer shopping preference prediction. In this way, the shopping preference of the user is accurately predicted by utilizing big data in the electronic commerce system, so that more personalized and intelligent commodity recommendation service is provided for the user.

Specifically, in the technical scheme of the application, firstly, relevant data of an analyzed user is extracted from an electronic commerce system, wherein the relevant data comprises a historical transaction record, search historical data and browsing historical data. Here, the history transaction record may reflect purchase habits, purchase power, and purchase frequency of the user, the search history data may reflect search intention, search keywords, and search results of the user, and the browsing history data may reflect browsing interests, browsing duration, and browsing contents of the user. Such data may help build a user representation.

In the technical scheme of the application, firstly, the user portrait of the analyzed user is constructed based on the related data of the analyzed user, wherein the user portrait comprises basic attributes, purchasing behavior, interests and purchasing intention. It should be appreciated that the user representation is a comprehensive description of the analyzed user that may reflect the personalized information of the analyzed user from multiple dimensions. In particular, the basic attributes are some fixed characteristics of the user, such as age, gender, region, occupation, etc., which may reflect the consumption level and preferences of the user; the purchasing behavior is the actual transaction condition of the user on the electronic commerce platform, such as purchasing frequency, purchasing amount, purchasing class and the like, which can reflect the consumption capability and habit of the user; the interest and hobbies are non-transaction behaviors of the user on the electronic commerce platform, such as searching keywords, browsing commodities, collecting commodities and the like, and can reflect the consumption interest and the demand of the user; the purchase intent is a potential transaction behavior of the user on the e-commerce platform, such as joining a shopping cart, click-to-settle, filling an order, etc., which may reflect the user's motivation and decisions for consumption. In this way, the consumer's mind and behavior may be more fully understood. And then, enabling each attribute item in the user portrait of the analyzed user to pass through a word embedding layer to obtain a plurality of attribute item word embedding vectors. Here, the word embedding layer may convert discrete words into a continuous vector representation. That is, each attribute item (basic attribute, purchase behavior, interest and purchase intention) in the user representation can be converted into a plurality of attribute item word embedding vectors through the word embedding layer, so that the operation and processing of the subsequent computer model are facilitated. And then, respectively adding the attribute values of all the attribute terms in the user portrait of the analyzed user to the tail parts of all the attribute term word embedded vectors to obtain a plurality of user portrait attribute term embedded vectors. By the method, the plurality of user portrait attribute word embedding vectors not only contain semantic information of attribute items, but also contain specific information of attribute values, so that the representation capability of the user portrait is improved.

Since each attribute term and attribute value in the user portrait attribute word embedded vector are meaningful semantic units, there may be complex semantic association and logical reasoning between them, for example, the user's age, gender, occupation, etc. basic attributes may affect the user's interests and buying intent, and the user's search history data and browsing history data may reflect the user's buying behavior and preferences, etc.

In order to mine the semantic association and the logical reasoning relation, in the technical scheme of the application, the plurality of user portrait attribute words are embedded into vectors to obtain first-scale user portrait semantic understanding feature vectors through a context encoder based on a converter. That is, the context encoder based on the converter can effectively perform global semantic understanding on the plurality of user portrait attribute word embedded vectors, and the output first-scale user portrait semantic understanding feature vector can express long-distance dependency semantic relations between each attribute item and attribute value in the user portrait. And meanwhile, arranging the plurality of user portrait attribute word embedded vectors into a two-dimensional feature matrix, and obtaining a second-scale user portrait semantic understanding feature vector through a text convolutional neural network model. Here, the text convolutional neural network model can extract local semantic information from text data. Specifically, the text convolutional neural network model captures semantic features within a local neighborhood by sliding window operations using convolutional collation text data.

In the technical scheme of the application, the context encoder based on the converter and the text convolutional neural network model capture implicit relations among user portrait attributes respectively in different scales. Further, the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector are fused to obtain a user portrait semantic understanding feature vector. In this way, the expressive power and generalization power of the user portrait semantic understanding feature vector are enhanced.

And then, the user portrait semantic understanding feature vector passes through a classifier to obtain a classification result, wherein the classification result is used for representing shopping preference type labels. Here, the classifier can learn a classification rule according to the corresponding relation between the user portrait semantic understanding feature vector and the shopping preference type label in the training data, so as to predict the inferred user portrait semantic understanding feature vector and output the corresponding shopping preference type label. It is worth mentioning that the shopping preference type tag is a simplification and generalization of shopping preferences of users, and can divide users into different categories, such as price sensitive type, brand loyalty type, quality pursuit type, etc., and in subsequent applications, can be modified according to actual situations.

In the technical scheme of the application, the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector respectively express text associated semantic features of each attribute item of the user portrait of the analyzed user under different scales, so that when the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector are fused to obtain the user portrait semantic understanding feature vector, in order to promote the fusion effect of the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector, the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector need to be fused on a feature semantic level.

Accordingly, applicants of the present application semantically understand feature vectors, e.g., denoted as V, for the first-scale user representation ₁ And said second scale user portrayal semantic understanding feature vector, e.g., denoted as V ₂ Performing deep space encapsulation semantic matching fusion to obtain the user portrait semantic understanding feature vector, for example, expressed as V _c Wherein the user portrayal semantic understanding feature vector V _c The method comprises the following steps:

||·|| ₁ and|| | ₂ Represents the first norm and the second norm of the vector respectively, ω and ε are the weight and bias hyper-parameters respectively, D (V ₁ ,V ₂ ) Representing the first scale user portrayal semantic understanding feature vector V ₁ And the second scale user portrayal semantic understanding feature vector V ₂ A per-position distance matrix between, i.e. the feature value of each position of the per-position distance matrix is the first scale user portrait semantic understanding feature vector V ₁ And the second scale user portrayal semantic understanding feature vector V ₂ The distance between the eigenvalues of the corresponding positions of (c), denoted D _ij ＝d(v _1i ,v _2j ) And I is an identity matrix.

Here, feature vector V is semantically understood for the first scale user representation in depth feature space ₁ And the second scale user portrayal semantic understanding feature vector V ₂ The semantic expression is packaged into a deep space, so that fine-grained features in the overall distribution of the feature vector simultaneously comprise low-level semantic distribution and high-level semantic distribution, thereby, through the deep space packaging semantic matching fusion, the matching of semantic levels of a classification mode layer can be performed through balancing the low-level semantic distribution and the high-level semantic distribution, so as to realize the semantic controlled compiling fusion of the features in the feature space, and further, the first-scale user portrait semantic understanding feature vector is obtained V ₁ And the second scale user portrayal semantic understanding feature vector V ₂ Semantic collaboration in feature fusion space improves semantic understanding feature vector V of the user portrait _c Semantic understanding feature vector V for the first scale user representation ₁ And the second scale user portrayal semantic understanding feature vector V ₂ The semantic fusion effect of the user portrait semantic understanding feature vector is improved, and therefore the accuracy of a classification result obtained by the user portrait semantic understanding feature vector through the classifier is improved.

Based on this, the application proposes an electronic commerce management system based on big data, which comprises: the system comprises a user data acquisition module, a user data analysis module and a user data analysis module, wherein the user data acquisition module is used for extracting relevant data of an analyzed user from an electronic commerce system, and the relevant data comprises a historical transaction record, search historical data and browsing historical data; a user portrayal construction module for constructing a user portrayal of the analyzed user based on the related data of the analyzed user, wherein the user portrayal comprises basic properties, purchasing behavior, interests and purchasing intention; the word embedding module is used for enabling each attribute item in the user portrait of the analyzed user to pass through a word embedding layer to obtain a plurality of attribute item word embedding vectors; the attribute value adding module is used for respectively adding the attribute value of each attribute term in the user portrait of the analyzed user to the tail part of each attribute term word embedding vector so as to obtain a plurality of user portrait attribute word embedding vectors; the first semantic understanding module is used for enabling the plurality of user portrait attribute words to be embedded into vectors and enabling the user portrait attribute words to pass through a context encoder based on a converter so as to obtain first-scale user portrait semantic understanding feature vectors; the second semantic understanding module is used for arranging the plurality of user portrait attribute word embedded vectors into a two-dimensional feature matrix and then obtaining second-scale user portrait semantic understanding feature vectors through a text convolutional neural network model; the feature fusion module is used for fusing the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector to obtain a user portrait semantic understanding feature vector; and the label dividing module is used for enabling the user portrait semantic understanding feature vector to pass through a classifier to obtain a classification result, and the classification result is used for representing shopping preference type labels.

Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.

Exemplary System

Fig. 1 is a block diagram of a big data based e-commerce management system in accordance with an embodiment of the present application. As shown in fig. 1, the big data based e-commerce management system 300 according to an embodiment of the present application includes: a user data acquisition module 310; a user portrayal construction module 320; a word embedding module 330; an attribute value adding module 340; a first semantic understanding module 350; a second semantic understanding module 360; a feature fusion module 370; and a tag dividing module 380.

Wherein the user data obtaining module 310 is configured to extract relevant data of the analyzed user from the e-commerce system, where the relevant data includes a historical transaction record, search historical data and browsing historical data; the user portrait construction module 320 is configured to construct a user portrait of the analyzed user based on the related data of the analyzed user, where the user portrait includes basic attributes, purchase behaviors, interests, and purchase intents; the word embedding module 330 is configured to pass each attribute term in the user representation of the analyzed user through a word embedding layer to obtain a plurality of attribute term word embedding vectors; the attribute value adding module 340 is configured to add attribute values of each attribute term in the user portrait of the analyzed user to tail portions of the attribute term word embedding vectors respectively so as to obtain a plurality of user portrait attribute term embedding vectors; the first semantic understanding module 350 is configured to insert the plurality of user portrait attribute words into vectors through a context encoder based on a converter to obtain first scale user portrait semantic understanding feature vectors; the second semantic understanding module 360 is configured to arrange the plurality of user portrait attribute word embedding vectors into a two-dimensional feature matrix, and obtain second-scale user portrait semantic understanding feature vectors through a text convolutional neural network model; the feature fusion module 370 is configured to fuse the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector to obtain a user portrait semantic understanding feature vector; and the label dividing module 380 is configured to pass the user portrait semantic understanding feature vector through a classifier to obtain a classification result, where the classification result is used to represent a shopping preference type label.

Fig. 2 is a system architecture diagram of a big data based e-commerce management system according to an embodiment of the present application. As shown in fig. 2, in the network architecture, related data of the analyzed user is first extracted from the e-commerce system through the user data acquisition module 310, wherein the related data includes a historical transaction record, search historical data and browsing historical data; next, the user portrayal construction module 320 constructs a user portrayal of the analyzed user based on the related data of the analyzed user acquired by the user data acquisition module 310, wherein the user portrayal includes basic attributes, purchase behavior, interests and purchase intention; the word embedding module 330 passes each attribute term in the user portrait of the analyzed user obtained by the user portrait construction module 320 through a word embedding layer to obtain a plurality of attribute term word embedding vectors; the attribute value adding module 340 adds the attribute value of each attribute term in the user portrait of the analyzed user to the tail of each attribute term word embedding vector so as to obtain a plurality of user portrait attribute term embedding vectors; then, the first semantic understanding module 350 passes the plurality of user portrait attribute word embedding vectors obtained by the attribute value adding module 340 through a context encoder based on a converter to obtain first scale user portrait semantic understanding feature vectors; the second semantic understanding module 360 arranges the plurality of user portrait attribute word embedded vectors obtained by the attribute value adding module 340 into a two-dimensional feature matrix, and then obtains second-scale user portrait semantic understanding feature vectors through a text convolutional neural network model; the feature fusion module 370 fuses the first-scale user portrait semantic understanding feature vector obtained by the first semantic understanding module 350 and the second-scale user portrait semantic understanding feature vector obtained by the second semantic understanding module 360 to obtain a user portrait semantic understanding feature vector; and the tag dividing module 380 enables the feature fusion module 370 to fuse the obtained user portrait semantic understanding feature vectors to pass through a classifier to obtain a classification result, wherein the classification result is used for representing shopping preference type tags.

Specifically, during the operation of the big data-based e-commerce management system 300, the user data obtaining module 310 is configured to extract relevant data of the analyzed user from the e-commerce system, where the relevant data includes a historical transaction record, search history data and browsing history data. In the technical scheme of the application, the aim of consumer shopping preference prediction is fulfilled by comprehensively analyzing the data. In one example, the historical transaction record may reflect a user's buying habit, buying power, and frequency of purchases, the search history data may reflect a user's search intent, search keywords, and search results, and the browsing history data may reflect a user's browsing interests, browsing duration, and browsing content. Such data may help build a user representation.

Specifically, during operation of the big data based e-commerce management system 300, the user portrayal construction module 320 is configured to construct a user portrayal of the analyzed user based on the relevant data of the analyzed user, where the user portrayal includes basic attributes, purchasing behavior, interests and purchasing intent. In the technical scheme of the application, firstly, the user portrait of the analyzed user is constructed based on the related data of the analyzed user, wherein the user portrait comprises basic attributes, purchasing behavior, interests and purchasing intention. It should be appreciated that the user representation is a comprehensive description of the analyzed user that may reflect the personalized information of the analyzed user from multiple dimensions. In particular, the basic attributes are some fixed characteristics of the user, such as age, gender, region, occupation, etc., which may reflect the consumption level and preferences of the user; the purchasing behavior is the actual transaction condition of the user on the electronic commerce platform, such as purchasing frequency, purchasing amount, purchasing class and the like, which can reflect the consumption capability and habit of the user; the interest and hobbies are non-transaction behaviors of the user on the electronic commerce platform, such as searching keywords, browsing commodities, collecting commodities and the like, and can reflect the consumption interest and the demand of the user; the purchase intent is a potential transaction behavior of the user on the e-commerce platform, such as joining a shopping cart, click-to-settle, filling an order, etc., which may reflect the user's motivation and decisions for consumption. In this way, the consumer's mind and behavior may be more fully understood.

Specifically, during the operation of the big data-based e-commerce management system 300, the word embedding module 330 and the attribute value adding module 340 are configured to pass each attribute term in the user representation of the analyzed user through a word embedding layer to obtain a plurality of attribute term word embedding vectors; and adding the attribute values of the attribute terms in the user portrait of the analyzed user to the tail parts of the attribute term word embedding vectors so as to obtain a plurality of user portrait attribute term embedding vectors. That is, each attribute term in the user representation can be converted into a plurality of attribute term word embedding vectors through the word embedding layer, so that the operation and the processing of the subsequent computer model are facilitated. In one specific example, first, each attribute item in a user representation of the analyzed user is expanded into a one-dimensional pixel vector to obtain a sequence of a plurality of one-dimensional pixel vectors; and then using the weight matrix of the embedding layer to carry out embedding coding on one-dimensional pixel vectors in the sequence of the plurality of one-dimensional pixel vectors so as to obtain the plurality of attribute term word embedding vectors. And after the plurality of attribute term word embedded vectors are obtained, respectively adding attribute values of all attribute terms in the user portrait of the analyzed user to the tail parts of all the attribute term word embedded vectors so as to obtain a plurality of user portrait attribute term embedded vectors. By the method, the plurality of user portrait attribute word embedding vectors not only contain semantic information of attribute items, but also contain specific information of attribute values, so that the representation capability of the user portrait is improved.

Specifically, during operation of the big data based e-commerce management system 300, the first semantic understanding module 350 is configured to insert the plurality of user portrait attribute words into vectors through a context encoder based on a converter to obtain first scale user portrait semantic understanding feature vectors. Considering that each attribute item and each attribute value in the user portrait attribute word embedding vector are meaningful semantic units, complex semantic association and logic reasoning can exist between the attribute items and the attribute values, therefore, in the technical scheme of the application, the plurality of user portrait attribute word embedding vectors pass through a context encoder based on a converter to obtain a first-scale user portrait semantic understanding feature vector. That is, the context encoder based on the converter can effectively perform global semantic understanding on the plurality of user portrait attribute word embedded vectors, and the output first-scale user portrait semantic understanding feature vector can express long-distance dependency semantic relations between each attribute item and attribute value in the user portrait.

Fig. 3 is a block diagram of a first semantic understanding module in a big data based e-commerce management system according to an embodiment of the present application. As shown in fig. 3, the first semantic understanding module 350 includes: a query vector construction unit 351, configured to perform one-dimensional arrangement on the plurality of user portrait attribute word embedded vectors to obtain a global user portrait feature vector; a self-attention unit 352 for calculating the product between the global user portrait character vector and the transpose vector of each of the plurality of user portrait character word embedded vectors to obtain a plurality of self-attention association matrices; a normalization unit 353, configured to perform normalization processing on each of the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices; a degree of attention calculation unit 354 configured to obtain a plurality of probability values by using a Softmax classification function for each normalized self-attention correlation matrix in the plurality of normalized self-attention correlation matrices; an attention applying unit 355, configured to weight each of the user portrait attribute word embedding vectors with each of the plurality of probability values as a weight to obtain the plurality of context semantic user portrait feature vectors; and a concatenation unit 356, configured to concatenate the plurality of context semantic user portrayal feature vectors to obtain the first-scale user portrayal semantic understanding feature vector.

Specifically, during the operation of the big data-based e-commerce management system 300, the second semantic understanding module 360 is configured to arrange the plurality of user portrait attribute words embedded vectors into a two-dimensional feature matrix, and then obtain second-scale user portrait semantic understanding feature vectors through a text convolutional neural network model. That is, a two-dimensional feature matrix obtained by arranging the plurality of user portrayal attribute word embedding vectors is convolutionally encoded using a text convolutional neural network model having excellent expression in terms of local semantic information extraction to perform a sliding window operation by collating text data using convolution, thereby capturing semantic features in local neighborhood. In one specific example, the text convolutional neural network includes a plurality of neural network layers that are cascaded with each other, wherein each neural network layer includes a convolutional layer, a pooling layer, and an activation layer. In the coding process of the text convolutional neural network, each layer of the text convolutional neural network carries out convolutional processing based on a convolutional kernel on input data by using the convolutional layer in the forward transmission process of the layer, carries out pooling processing on a convolutional feature map output by the convolutional layer by using the pooling layer and carries out activation processing on the pooled feature map output by the pooling layer by using the activation layer.

Fig. 4 is a flowchart of text convolutional neural network coding in a big data based e-commerce management system according to an embodiment of the present application. As shown in fig. 4, in the text convolutional neural network coding process, the method includes: each layer of the text convolutional neural network model is used for respectively carrying out forward transfer on input data in the layers: s210, carrying out convolution processing on input data to obtain a convolution characteristic diagram; s220, pooling the convolution feature map based on a feature matrix to obtain a pooled feature map; s230, carrying out nonlinear activation on the pooled feature map to obtain an activated feature map; the output of the last layer of the text convolutional neural network is the semantic understanding feature vector of the second-scale user portrait, and the input of the first layer of the text convolutional neural network is a two-dimensional feature matrix obtained by arranging the attribute word embedded vectors of the plurality of user portrait.

Specifically, during the operation of the big data-based e-commerce management system 300, the feature fusion module 370 is configured to fuse the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector to obtain a user portrait semantic understanding feature vector. That is, after the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector are obtained, features of the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector are further fused, so that implicit relations among user portrait attributes are captured in different scales. The first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector respectively express text associated semantic features of all attribute items of the user portrait of the analyzed user under different scales, so that when the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector are fused to obtain the user portrait semantic understanding feature vector, in order to promote the fusion effect of the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector, the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector need to be fused on a feature semantic level. Accordingly, applicants of the present application semantically understand feature vectors, e.g., denoted as V, for the first-scale user representation ₁ And said second scale user portrayal semantic understanding feature vector, e.g., denoted as V ₂ Performing deep space encapsulation semantic matching fusion to obtain the user portrait semantic understanding feature vector, for example, expressed as V _c Wherein the user portrayal semantic understanding feature vector V _c The method comprises the following steps:

||·|| ₁ and|| | ₂ Represents the first norm and the second norm of the vector respectively, ω and ε are the weight and bias hyper-parameters respectively, D (V ₁ ,V ₂ ) Representing the first scale user portrayal semantic understanding feature vector V ₁ And the second scale user portrayal semantic understanding feature vector V ₂ A per-position distance matrix between, i.e. the feature value of each position of the per-position distance matrix is the first scale user portrait semantic understanding feature vector V ₁ And the second scale user portrayal semantic understanding feature vector V ₂ The distance between the eigenvalues of the corresponding positions of (c), denoted D _ij ＝d(v _1i ,v _2j ) And I is an identity matrix. Here, feature vector V is semantically understood for the first scale user representation in depth feature space ₁ And the second scale user portrayal semantic understanding feature vector V ₂ The semantic expression is packaged into a deep space, so that fine-grained features in the overall distribution of the feature vector simultaneously comprise low-level semantic distribution and high-level semantic distribution, thereby, through the deep space packaging semantic matching fusion, the matching of semantic levels of a classification mode layer can be performed through balancing the low-level semantic distribution and the high-level semantic distribution, so as to realize the semantic controlled compiling fusion of the features in the feature space, and further, the first-scale user portrait semantic understanding feature vector V is obtained ₁ And the second scale user portrayal semantic understanding feature vector V ₂ Semantic collaboration in feature fusion space improves semantic understanding feature vector V of the user portrait _c Semantic understanding feature vector V for the first scale user representation ₁ And the second scale user portrayal semantic understanding feature vector V ₂ The semantic fusion effect of the user portrait semantic understanding feature vector is improved, so that the classification result of the user portrait semantic understanding feature vector obtained by a classifier is improvedAccuracy of (3).

Specifically, during the operation of the big data-based e-commerce management system 300, the tag classification module 380 is configured to pass the user portrait semantic understanding feature vector through a classifier to obtain a classification result, where the classification result is used to represent a shopping preference type tag. In the technical scheme of the application, after the user portrait semantic understanding feature vector is obtained, the user portrait semantic understanding feature vector is further used as a classification feature vector to be subjected to classification processing through a classifier so as to obtain a classification result for representing shopping preference type labels. Specifically, the classifier includes a plurality of fully connected layers and a Softmax layer cascaded with a last fully connected layer of the plurality of fully connected layers. In the classification processing of the classifier, a plurality of full-connection layers of the classifier are used for carrying out full-connection coding on the user portrait semantic understanding feature vector so as to obtain a coded user portrait semantic understanding feature vector; further, the encoded user portrait semantic understanding feature vector is input into a Softmax layer of the classifier, i.e. classification processing is performed on the encoded user portrait semantic understanding feature vector by using the Softmax classification function to obtain a classification label. Here, the classifier can learn a classification rule according to the corresponding relation between the user portrait semantic understanding feature vector and the shopping preference type label in the training data, so as to predict the inferred user portrait semantic understanding feature vector and output the corresponding shopping preference type label. It is worth mentioning that the shopping preference type tag is a simplification and generalization of shopping preferences of users, and can divide users into different categories, such as price sensitive type, brand loyalty type, quality pursuit type, etc., and in subsequent applications, can be modified according to actual situations.

In summary, a big data based e-commerce management system 300 is illustrated that achieves the objective of consumer shopping preference prediction by comprehensively utilizing historical transaction records, search history data and browsing history data using deep learning based artificial intelligence techniques. In this way, the shopping preference of the user is accurately predicted by utilizing big data in the electronic commerce system, so that more personalized and intelligent commodity recommendation service is provided for the user.

As described above, the big data based e-commerce management system according to the embodiment of the present application can be implemented in various terminal devices. In one example, big data based e-commerce management system 300 according to embodiments of the present application may be integrated into a terminal device as a software module and/or hardware module. For example, the big data based e-commerce management system 300 may be a software module in the operating system of the terminal device or may be an application developed for the terminal device; of course, the big data based e-commerce management system 300 may also be one of a plurality of hardware modules of the terminal device.

Alternatively, in another example, the big data based e-commerce management system 300 and the terminal device may be separate devices, and the big data based e-commerce management system 300 may be connected to the terminal device through a wired and/or wireless network and transmit the interactive information in a agreed data format.

Exemplary method

Fig. 5 is a flowchart of an electronic commerce management method based on big data according to an embodiment of the present application. As shown in fig. 5, the big data based e-commerce management method according to the embodiment of the present application includes the steps of: s110, extracting relevant data of the analyzed user from the electronic commerce system, wherein the relevant data comprises historical transaction records, search historical data and browsing historical data; s120, constructing a user portrait of the analyzed user based on the related data of the analyzed user, wherein the user portrait comprises basic attributes, purchasing behavior, interests and purchasing intention; s130, enabling each attribute item in the user portrait of the analyzed user to pass through a word embedding layer to obtain a plurality of attribute item word embedding vectors; s140, respectively adding the attribute values of all attribute terms in the user portrait of the analyzed user to the tail parts of all attribute term embedding vectors to obtain a plurality of user portrait attribute term embedding vectors; s150, enabling the plurality of user portrait attribute word embedded vectors to pass through a context encoder based on a converter to obtain first-scale user portrait semantic understanding feature vectors; s160, arranging the embedded vectors of the user portrait attribute words into a two-dimensional feature matrix, and obtaining semantic understanding feature vectors of the user portrait of a second scale through a text convolutional neural network model; s170, fusing the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector to obtain a user portrait semantic understanding feature vector; and S180, enabling the user portrait semantic understanding feature vector to pass through a classifier to obtain a classification result, wherein the classification result is used for representing shopping preference type labels.

In one example, in the electronic commerce management method based on big data, the step S130 includes: expanding each attribute term in the user representation of the analyzed user into a one-dimensional pixel vector to obtain a sequence of a plurality of one-dimensional pixel vectors; and performing embedded coding on one-dimensional pixel vectors in the sequence of the plurality of one-dimensional pixel vectors by using the weight matrix of the embedded layer to obtain the plurality of attribute term embedded vectors.

In one example, in the electronic commerce management method based on big data, the step S150 includes: one-dimensional arrangement is carried out on the embedding vectors of the user portrait attribute words so as to obtain global user portrait feature vectors; calculating the product between the global user portrait character vector and the transpose vector of each user portrait character word embedding vector in the plurality of user portrait character word embedding vectors to obtain a plurality of self-attention association matrices; respectively carrying out standardization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of standardized self-attention correlation matrices; obtaining a plurality of probability values by using a Softmax classification function through each normalized self-attention correlation matrix in the normalized self-attention correlation matrices; weighting each user portrait attribute word embedded vector in the plurality of user portrait attribute word embedded vectors by taking each probability value in the plurality of probability values as a weight so as to obtain the plurality of context semantic user portrait feature vectors; and cascading the plurality of context semantic user portrayal feature vectors to obtain the first-scale user portrayal semantic understanding feature vector.

In one example, in the electronic commerce management method based on big data, the step S160 includes: each layer of the text convolutional neural network model is used for respectively carrying out forward transfer on input data in the layers: carrying out convolution processing on input data to obtain a convolution characteristic diagram; pooling the convolution feature images based on a feature matrix to obtain pooled feature images; performing nonlinear activation on the pooled feature map to obtain an activated feature map; the output of the last layer of the text convolutional neural network is the semantic understanding feature vector of the second-scale user portrait, and the input of the first layer of the text convolutional neural network is a two-dimensional feature matrix obtained by arranging the attribute word embedded vectors of the plurality of user portrait.

In one example, in the electronic commerce management method based on big data, the step S170 includes: carrying out deep space packaging semantic matching fusion on the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector by using the following formula to obtain the user portrait semantic understanding feature vector; wherein, the formula is:

In one example, in the electronic commerce management method based on big data, the step S180 includes: performing full-connection coding on the user portrait semantic understanding feature vector by using a plurality of full-connection layers of the classifier to obtain a coded user portrait semantic understanding feature vector; and passing the encoded user portrayal semantic understanding feature vector through a Softmax classification function of the classifier to obtain the classification result.

In summary, the big data-based e-commerce management method according to the embodiment of the present application is explained, which achieves the purpose of consumer shopping preference prediction by comprehensively utilizing historical transaction records, search history data and browsing history data by adopting artificial intelligence technology based on deep learning. In this way, the shopping preference of the user is accurately predicted by utilizing big data in the electronic commerce system, so that more personalized and intelligent commodity recommendation service is provided for the user.

Exemplary electronic device

Next, an electronic device according to an embodiment of the present application is described with reference to fig. 6.

Fig. 6 illustrates a block diagram of an electronic device according to an embodiment of the application.

As shown in fig. 6, the electronic device 10 includes one or more processors 11 and a memory 12.

The processor 11 may be a Central Processing Unit (CPU) or other form of processing unit having data processing and/or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

Memory 12 may include one or more computer program products that may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. The volatile memory may include, for example, random Access Memory (RAM) and/or cache memory (cache), and the like. The non-volatile memory may include, for example, read Only Memory (ROM), hard disk, flash memory, and the like. One or more computer program instructions may be stored on the computer readable storage medium that may be executed by the processor 11 to implement the functions in the big data based e-commerce management system and/or other desired functions of the various embodiments of the present application described above. Various content such as user portrait semantic understanding feature vectors may also be stored in the computer readable storage medium.

In one example, the electronic device 10 may further include: an input device 13 and an output device 14, which are interconnected by a bus system and/or other forms of connection mechanisms (not shown).

The input means 13 may comprise, for example, a keyboard, a mouse, etc.

The output device 14 may output various information including the classification result and the like to the outside. The output means 14 may include, for example, a display, speakers, a printer, and a communication network and remote output devices connected thereto, etc.

Of course, only some of the components of the electronic device 10 that are relevant to the present application are shown in fig. 6 for simplicity, components such as buses, input/output interfaces, etc. are omitted. In addition, the electronic device 10 may include any other suitable components depending on the particular application.

Exemplary computer program product and computer readable storage Medium

In addition to the methods and apparatus described above, embodiments of the application may also be a computer program product comprising computer program instructions which, when executed by a processor, cause the processor to perform steps in the functions of the big data based e-commerce management method according to the various embodiments of the application described in the "exemplary systems" section of this specification.

The computer program product may write program code for performing operations of embodiments of the present application in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server.

Furthermore, embodiments of the present application may also be a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, cause the processor to perform steps in the functions of the big data based e-commerce management method according to the various embodiments of the present application described in the "exemplary systems" section above of the present specification.

The computer readable storage medium may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The basic principles of the present application have been described above in connection with specific embodiments, however, it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are merely examples and not intended to be limiting, and these advantages, benefits, effects, etc. are not to be considered as essential to the various embodiments of the present application. Furthermore, the specific details disclosed herein are for purposes of illustration and understanding only, and are not intended to be limiting, as the application is not necessarily limited to practice with the above described specific details.

The block diagrams of the devices, apparatuses, devices, systems referred to in the present application are only illustrative examples and are not intended to require or imply that the connections, arrangements, configurations must be made in the manner shown in the block diagrams. As will be appreciated by one of skill in the art, the devices, apparatuses, devices, systems may be connected, arranged, configured in any manner. Words such as "including," "comprising," "having," and the like are words of openness and mean "including but not limited to," and are used interchangeably therewith. The terms "or" and "as used herein refer to and are used interchangeably with the term" and/or "unless the context clearly indicates otherwise. The term "such as" as used herein refers to, and is used interchangeably with, the phrase "such as, but not limited to.

It is also noted that in the apparatus, devices and methods of the present application, the components or steps may be disassembled and/or assembled. Such decomposition and/or recombination should be considered as equivalent aspects of the present application.

The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

The foregoing description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit embodiments of the application to the form disclosed herein. Although a number of example aspects and embodiments have been discussed above, a person of ordinary skill in the art will recognize certain variations, modifications, alterations, additions, and subcombinations thereof.

Claims

1. An electronic commerce management system based on big data, comprising:

2. The big data based e-commerce management system of claim 1, wherein the word embedding module comprises:

a one-dimensional unfolding unit for unfolding each attribute item in the user portrait of the analyzed user into a one-dimensional pixel vector to obtain a sequence of a plurality of one-dimensional pixel vectors; and

and the embedding encoding unit is used for carrying out embedding encoding on one-dimensional pixel vectors in the sequence of the plurality of one-dimensional pixel vectors by using the weight matrix of the embedding layer so as to obtain the plurality of attribute term word embedding vectors.

3. The big data based e-commerce management system of claim 2, wherein the first semantic understanding module comprises:

the query vector construction unit is used for carrying out one-dimensional arrangement on the plurality of user portrait attribute word embedded vectors so as to obtain global user portrait feature vectors;

a self-attention unit, configured to calculate a product between the global user portrait feature vector and a transpose vector of each user portrait attribute word embedding vector in the plurality of user portrait attribute word embedding vectors to obtain a plurality of self-attention correlation matrices;

The normalization unit is used for respectively performing normalization processing on each self-attention correlation matrix in the plurality of self-attention correlation matrices to obtain a plurality of normalized self-attention correlation matrices;

the attention calculating unit is used for obtaining a plurality of probability values through a Softmax classification function by each normalized self-attention correlation matrix in the normalized self-attention correlation matrices;

the attention applying unit is used for weighting each user portrait attribute word embedded vector in the plurality of user portrait attribute word embedded vectors by taking each probability value in the plurality of probability values as a weight so as to obtain the plurality of context semantic user portrait feature vectors; and

and the cascading unit is used for cascading the plurality of context semantic user portrait feature vectors to obtain the first-scale user portrait semantic understanding feature vector.

4. The big data based e-commerce management system of claim 3, wherein the second semantic understanding module comprises: each layer of the text convolutional neural network model is used for respectively carrying out forward transfer on input data in the layers:

carrying out convolution processing on input data to obtain a convolution characteristic diagram;

Pooling the convolution feature images based on a feature matrix to obtain pooled feature images; and

non-linear activation is carried out on the pooled feature map so as to obtain an activated feature map;

the output of the last layer of the text convolutional neural network is the semantic understanding feature vector of the second-scale user portrait, and the input of the first layer of the text convolutional neural network is a two-dimensional feature matrix obtained by arranging the attribute word embedded vectors of the plurality of user portrait.

5. The big data based e-commerce management system of claim 4, wherein the feature fusion module is further configured to:

carrying out deep space packaging semantic matching fusion on the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector by using the following formula to obtain the user portrait semantic understanding feature vector;

wherein, the formula is:

6. The big data based e-commerce management system of claim 5, wherein the tag partitioning module comprises:

the full-connection coding unit is used for carrying out full-connection coding on the user portrait semantic understanding feature vector by using a plurality of full-connection layers of the classifier so as to obtain a coded user portrait semantic understanding feature vector; and

and the classification result generating unit is used for enabling the encoded user portrait semantic understanding feature vector to pass through a Softmax classification function of the classifier to obtain the classification result.

7. An electronic commerce management method based on big data, comprising the following steps:

8. The method for electronic commerce management based on big data according to claim 7, wherein the step of arranging the plurality of user portrayal attribute word embedding vectors into a two-dimensional feature matrix and then obtaining second-scale user portrayal semantic understanding feature vectors through a text convolutional neural network model comprises: each layer of the text convolutional neural network model is used for respectively carrying out forward transfer on input data in the layers:

9. The big data based e-commerce management method of claim 8, wherein fusing the first scale user portrait semantic understanding feature vector and the second scale user portrait semantic understanding feature vector to obtain a user portrait semantic understanding feature vector comprises: carrying out deep space packaging semantic matching fusion on the first-scale user portrait semantic understanding feature vector and the second-scale user portrait semantic understanding feature vector by using the following formula to obtain the user portrait semantic understanding feature vector;

wherein, the formula is:

10. The big data based e-commerce management method of claim 9, wherein passing the user portrayal semantic understanding feature vector through a classifier to obtain a classification result, the classification result being used to represent a shopping preference type tag, comprises:

performing full-connection coding on the user portrait semantic understanding feature vector by using a plurality of full-connection layers of the classifier to obtain a coded user portrait semantic understanding feature vector; and

and passing the encoded user portrait semantic understanding feature vector through a Softmax classification function of the classifier to obtain the classification result.