CN117217875A

CN117217875A - Commodity recommendation method and device, electronic equipment and storage medium

Info

Publication number: CN117217875A
Application number: CN202311416533.3A
Authority: CN
Inventors: 蒋兴辉
Original assignee: Tianyi Electronic Commerce Co Ltd
Current assignee: Tianyi Electronic Commerce Co Ltd
Priority date: 2023-10-27
Filing date: 2023-10-27
Publication date: 2023-12-12

Abstract

The invention discloses a commodity recommendation method and device, electronic equipment and storage media, and relates to the technical field of electronic commerce, wherein the commodity recommendation method comprises the following steps: responding to a commodity searching request of a user side to obtain a user identification of the user side; acquiring a historical behavior data table of a user terminal based on a user identifier, and acquiring a commodity scoring table and a commodity dictionary table; building a commodity correlation characteristic table based on the historical behavior data table and the commodity scoring table, and building a commodity characteristic weight table based on the historical behavior data table and the commodity dictionary table; and calculating and sequencing weight values of the commodities in the commodity association characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list. The invention solves the technical problem that in the related technology, the user has poor online shopping experience caused by recommending interested products to the user side by calculating the similarity between the purchased goods and other goods in the mall.

Description

Commodity recommendation method and device, electronic equipment and storage medium

Technical Field

The invention relates to the technical field of electronic commerce, in particular to a commodity recommendation method and device, electronic equipment and a storage medium.

Background

With the rapid rise of each large e-commerce platform, online shopping becomes the current most mainstream shopping mode, with the gradual maturity of the shopping platform, the demand of users for commodity searching is also higher and higher, and in the online shopping scene, intelligent recommendation is still in a sprouting stage at present, similar products are recommended to users based on the shopping behaviors or searching behaviors of the users, so that the shopping demands of the users are met, products with more fitting demands are provided for the users, and the shopping interests of the users can be promoted.

In the related art, a similarity calculation mode is adopted to recommend personalized commodities to a user, namely, products of the same type after purchase or ordering are recommended to the user, so that the shopping requirements of multiple aspects of the user are met, meanwhile, for a new user, commodity recommendation cannot be effectively performed to the user under the condition that the purchase history and ordering history of the user cannot be acquired, and the problem of cold start exists.

In view of the above problems, no effective solution has been proposed at present.

Disclosure of Invention

The embodiment of the invention provides a commodity recommending method and device, electronic equipment and storage medium, and aims to at least solve the technical problem that in the related art, the user side is recommended to interested products by calculating the similarity between purchased commodities and other commodities in a mall, so that the online shopping experience of the user is poor.

According to an aspect of an embodiment of the present invention, there is provided a commodity recommendation method including: responding to a commodity searching request of a user side to obtain a user identification of the user side; acquiring a historical behavior data table of the user terminal based on the user identifier, and acquiring a commodity scoring table and a commodity dictionary table, wherein the historical behavior data table at least comprises: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: the commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics; constructing a commodity correlation feature table based on the historical behavior data table and the commodity scoring table, and constructing a commodity feature weight table based on the historical behavior data table and the commodity dictionary table; and calculating and sequencing weight values of the commodities in the commodity association characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list.

Optionally, after generating the commodity recommendation list based on the commodity list, the method further comprises: determining commodities to be recommended based on the commodity recommendation list; arranging the commodities to be recommended, rendering the terminal page of the user terminal based on the commodities to be recommended, and displaying the commodities to be recommended to the terminal page of the user terminal.

Optionally, the step of obtaining the historical behavior data table of the user terminal based on the user identifier includes: collecting shopping behavior data of the user terminal based on the user identifier, wherein the shopping behavior data at least comprises: the commodity identification and behavior characteristics; configuring a label value for each commodity based on the behavior characteristics to obtain a commodity label value; and generating the historical behavior data table of the user terminal based on the user identification, the commodity identification and the commodity label value.

Optionally, the step of obtaining the commodity dictionary table includes: information acquisition is carried out on each commodity in the mall to obtain basic commodity information, wherein the basic commodity information at least comprises: commodity identification and commodity characteristics; and assigning a weight value to each commodity feature to obtain the feature weight, and generating the commodity dictionary table based on the commodity identification, the commodity feature and the feature weight.

Optionally, the step of constructing a commodity correlation feature table based on the historical behavior data table and the commodity scoring table includes: combining each commodity in the commodity scoring table with each commodity in the historical behavior data table to obtain a set of associated commodity pairs; calculating a weight value of each associated commodity pair in the associated commodity pair set; calculating the association weight value of each commodity in the commodity scoring table and the user side based on the weight value of the associated commodity pair; and generating the commodity association characteristic table based on the association weight value of each commodity and the user side.

Optionally, the scoring model is pre-constructed, and the step of constructing the scoring model includes: acquiring commodity information of each commodity in a mall, wherein the commodity information at least comprises: commodity feature identification and commodity feature scoring values; constructing sample data based on the commodity information, and dividing the sample data into a training set and a testing set based on a preset proportion; training the classifier based on the training set to obtain an initial scoring model; and testing the initial scoring model based on the test set to obtain test accuracy, and obtaining the scoring model under the condition that the test accuracy is greater than or equal to a preset accuracy threshold.

Optionally, the step of constructing sample data based on the commodity information includes: preprocessing the commodity information, wherein the preprocessing at least comprises the following steps: desensitizing treatment, data cleaning treatment and data missing value treatment; and converting the preprocessed commodity information into vector data to obtain a commodity vector data set, and constructing the sample data based on the commodity vector data set.

According to another aspect of the embodiment of the present invention, there is also provided a commodity recommendation device, including: the response unit is used for responding to the commodity searching request of the user side and obtaining the user identification of the user side; the acquiring unit is configured to acquire a historical behavior data table of the user terminal based on the user identifier, and acquire a commodity scoring table and a commodity dictionary table, where the historical behavior data table at least includes: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: the commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics; the construction unit is used for constructing a commodity correlation characteristic table based on the historical behavior data table and the commodity scoring table, and constructing a commodity characteristic weight table based on the historical behavior data table and the commodity dictionary table; and the generation unit is used for calculating and sequencing the weight values of the commodities in the commodity association characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list.

Optionally, the commodity recommendation device further includes: the first determining module is used for determining the commodity to be recommended based on the commodity recommendation list; the first display module is used for arranging the commodities to be recommended, rendering the terminal page of the user terminal based on the commodities to be recommended, and displaying the commodities to be recommended to the terminal page of the user terminal.

Optionally, the acquiring unit includes: the first acquisition module is used for acquiring shopping behavior data of the user side based on the user identifier, wherein the shopping behavior data at least comprises: the commodity identification and behavior characteristics; the first configuration module is used for configuring a label value for each commodity based on the behavior characteristics to obtain a commodity label value; the first generation module is used for generating the historical behavior data table of the user terminal based on the user identification, the commodity identification and the commodity label value.

Optionally, the acquiring unit further includes: the second acquisition module is used for acquiring information of each commodity in the mall to obtain basic commodity information, wherein the basic commodity information at least comprises: commodity identification and commodity characteristics; and the first giving module is used for giving a weight value to each commodity feature to obtain the feature weight, and generating the commodity dictionary table based on the commodity identification, the commodity feature and the feature weight.

Optionally, the building unit comprises: the first combination module is used for combining each commodity in the commodity scoring table with each commodity in the historical behavior data table to obtain a related commodity pair set; the first calculation module is used for calculating the weight value of each associated commodity pair in the associated commodity pair set; the first calculation module is used for calculating the association weight value between each commodity in the commodity scoring table and the user side based on the weight value of the associated commodity pair; and the second generation module is used for generating the commodity association characteristic table based on the association weight value of each commodity and the user side.

Optionally, the commodity recommendation device further includes: the third acquisition module is used for acquiring commodity information of each commodity in the mall, wherein the commodity information at least comprises: commodity feature identification and commodity feature scoring values; the first construction module is used for constructing sample data based on the commodity information and dividing the sample data into a training set and a testing set based on a preset proportion; the first training module is used for training the classifier based on the training set to obtain an initial scoring model; the first testing module is used for testing the initial scoring model based on the testing set to obtain testing accuracy, and the scoring model is obtained under the condition that the testing accuracy is greater than or equal to a preset accuracy threshold.

Optionally, the first building module includes: the first preprocessing sub-module is used for preprocessing the commodity information, wherein the preprocessing at least comprises the following steps: desensitizing treatment, data cleaning treatment and data missing value treatment; the first conversion sub-module is used for converting the preprocessed commodity information into vector data to obtain a commodity vector data set, and constructing the sample data based on the commodity vector data set.

According to another aspect of the embodiment of the present invention, there is further provided a computer readable storage medium, where the computer readable storage medium includes a stored computer program, and when the computer program runs, the apparatus where the computer readable storage medium is controlled to execute any one of the commodity recommendation methods described above.

According to another aspect of the embodiments of the present invention, there is also provided an electronic device including one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement any one of the commodity recommendation methods described above.

In the present disclosure, the method comprises the following steps: firstly responding to a commodity searching request of a user side to obtain a user identifier of the user side; and acquiring a historical behavior data table of the user terminal based on the user identifier, and acquiring a commodity scoring table and a commodity dictionary table, wherein the historical behavior data table at least comprises: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics; then, a commodity correlation characteristic table is built based on the historical behavior data table and the commodity scoring table, and a commodity characteristic weight table is built based on the historical behavior data table and the commodity dictionary table; and finally, calculating and sequencing weight values of the commodities in the commodity association characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list.

In the method, the historical behavior data table of the user side, the commodity scoring table and the commodity dictionary table are used for comprehensively evaluating commodities in the mall, the commodity correlation characteristic table is constructed according to the historical behavior data table and the commodity scoring table, the commodities in the mall are correlated with historical commodities purchased by the user side, the interested degree of the commodities is determined, meanwhile, the commodity characteristic weight table is constructed according to the historical behavior data table and the commodity dictionary table, the characteristic weight of the commodities is determined, the weight value corresponding to the commodities is obtained, finally, the weight value calculation and the sorting are carried out according to the commodity correlation characteristic table and the commodity characteristic weight table, the commodity recommendation list corresponding to the user side is obtained, the demand degree of the user for the commodity is comprehensively analyzed, the commodity is accurately recommended to the user side, the satisfaction degree of the user is improved, and the technical problem that in-interest products are recommended to the user side through calculating the similarity between the purchased commodities and other commodities in the mall in the related art is solved, and the online shopping experience is poor is caused.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute a limitation on the application. In the drawings:

FIG. 1 is a flow chart of an alternative merchandise recommendation method according to an embodiment of the application;

FIG. 2 is a schematic diagram of an alternative merchandise recommendation flow according to an embodiment of the application;

FIG. 3 is a flow chart of an alternative online ordering of items according to an embodiment of the application;

FIG. 4 is a schematic diagram of an alternative merchandise recommendation apparatus according to an embodiment of the application;

fig. 5 is a block diagram of a hardware structure of an electronic device (or mobile device) of a commodity recommendation method according to an embodiment of the present application.

Detailed Description

In order that those skilled in the art will better understand the present application, a technical solution in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, shall fall within the scope of the present application.

It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

To facilitate an understanding of the invention by those skilled in the art, some terms or nouns involved in the various embodiments of the invention are explained below:

the support vector machine, support Vector Machine, SVM for short, is a machine learning method for classification and regression analysis.

Gradient boosting decision tree, gradient Boosting Decision Tree, GBDT for short, is a common integrated learning method, and the performance of the overall model is improved by iteratively training and combining decision tree models.

Random Forest is a common integrated learning method.

Gradient lifting framework Gradient Boosting Framework, XGBoost for short, is an enhanced decision tree algorithm.

ETL (extraction, conversion, loading) is a common data processing process, and is used for extracting data from a source system, and loading the data into a target system after conversion so as to meet different service requirements.

The data structure server, redis, is an open source memory data structure storage system that supports a variety of data structures, such as strings, hash tables, lists, collections, and the like.

It should be noted that, the commodity recommendation method and the device thereof in the present disclosure may be used in the technical field of electronic commerce to comprehensively evaluate a commodity and recommend the commodity according to a user requirement, and may also be used in any field other than the technical field of electronic commerce to comprehensively evaluate the commodity and recommend the commodity according to the user requirement.

It should be noted that, related information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present disclosure are information and data authorized by a user or sufficiently authorized by each party, and the collection, use and processing of related data need to comply with related laws and regulations and standards of related areas, and are provided with corresponding operation entries for the user to select authorization or rejection. For example, an interface is provided between the system and the relevant user or institution, before acquiring the relevant information, the system needs to send an acquisition request to the user or institution through the interface, and acquire the relevant information after receiving the consent information fed back by the user or institution.

The following embodiments of the present invention are applicable to various commodity recommendation systems/applications/devices. According to the invention, the historical behavior data table, the commodity scoring table and the commodity dictionary table at the user side are used for comprehensively evaluating commodities, a scoring model is established in advance for scoring the commodities, commodities required by the user are obtained, the shopping experience of the user is improved by taking the user as a center, trouble brought by information redundancy to the user is reduced, the shopping requirement of the user is met, and the opportunity cost of user selection is reduced.

Aiming at the cold start problem, the invention provides the method for acquiring the comprehensive weight of each commodity through the feature index, thereby ensuring that the public preference commodity can be recommended for the user when the new user logs in Shang Ping, and solving the cold start problem.

The present invention will be described in detail with reference to the following examples.

Example 1

In accordance with an embodiment of the present invention, there is provided an embodiment of a commodity recommendation method, it being noted that the steps shown in the flowchart of the accompanying drawings may be performed in a computer system, such as a set of computer executable instructions, and, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in an order other than that shown or described herein.

FIG. 1 is a flow chart of an alternative merchandise recommendation method according to an embodiment of the invention, as shown in FIG. 1, comprising the steps of:

step S101, responding to a commodity searching request of a user side to obtain a user identification of the user side;

step S102, a historical behavior data table of a user side is obtained based on a user identifier, and a commodity scoring table and a commodity dictionary table are obtained, wherein the historical behavior data table at least comprises: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics;

step S103, constructing a commodity correlation characteristic table based on the historical behavior data table and the commodity scoring table, and constructing a commodity characteristic weight table based on the historical behavior data table and the commodity dictionary table;

and step S104, calculating and sequencing weight values of the commodities in the commodity correlation characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list.

Through the steps, firstly responding to a commodity searching request of a user side to obtain a user identifier of the user side; and acquiring a historical behavior data table of the user terminal based on the user identifier, and acquiring a commodity scoring table and a commodity dictionary table, wherein the historical behavior data table at least comprises: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics; then, a commodity correlation characteristic table is built based on the historical behavior data table and the commodity scoring table, and a commodity characteristic weight table is built based on the historical behavior data table and the commodity dictionary table; and finally, calculating and sequencing weight values of the commodities in the commodity association characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list.

In the embodiment, the historical behavior data table, the commodity scoring table and the commodity dictionary table of the user side are used for comprehensively evaluating commodities in the mall, the commodity correlation characteristic table is constructed according to the historical behavior data table and the commodity scoring table, the commodities in the mall are correlated with historical commodities purchased by the user side, the interested degree of the commodities is determined, meanwhile, the commodity characteristic weight table is constructed according to the historical behavior data table and the commodity dictionary table, the characteristic weight of the commodities is determined, the weight value corresponding to the commodities is obtained, finally, the weight value calculation and the sorting are carried out according to the commodity correlation characteristic table and the commodity characteristic weight table, the commodity recommendation list corresponding to the user side is obtained, the demand degree of the user on the commodities is comprehensively analyzed, the demand commodity is accurately recommended to the user side, the satisfaction degree of the user is improved, and the technical problem that in-interested products are recommended to the user side through calculating the similarity between the purchased commodities and other commodities in the mall in the related technology is solved, and the online shopping experience is poor is caused.

Embodiments of the present invention will be described in detail with reference to the following steps.

It should be noted that, the implementation subject of the embodiment of the present invention is a commodity recommendation system, which may be deployed on each e-commerce platform, and obtains user preference and demand through the commodity recommendation system, so as to recommend personalized commodities for users, and promote interests of online shopping for users.

It should be noted that, in the existing commodity recommendation technology, similarity calculation is performed on commodities purchased by a user or browsed commodities and other commodities in a mall, and recommendation is performed on the commodities for the user according to the similarity value obtained by calculation, but a large amount of similar commodities are recommended according to the purchasing behavior of the user, so that the user passively accepts the purchased similar commodities on an e-commerce platform, and is easy to produce burnout, lack of freshness, weak in purchasing wish and motivation, and the existing recommendation mode belongs to large data commodity recommendation.

In addition, the existing commodity recommendation technology has the problem of cold start, no behavior record is generated under the condition that a new user enters a platform, the platform cannot acquire the user behavior data and the user characteristic data, cannot acquire commodities required by the user according to the user history behaviors, cannot meet the preference of the user, is easy to cause user loss, meanwhile, the system automatically scores the commodities according to the stay time for the situation that the user temporarily leaves a browsing and paying page, is easy to cause misjudgment, and other special situations such as malicious bad evaluation of competitors and the like are also possible to cause misjudgment.

Optionally, the scoring model is pre-constructed, and the step of constructing the scoring model includes: acquiring commodity information of each commodity in the mall, wherein the commodity information at least comprises: commodity feature identification and commodity feature scoring values; constructing sample data based on commodity information, and dividing the sample data into a training set and a testing set based on a preset proportion; training the classifier based on the training set to obtain an initial scoring model; and testing the initial scoring model based on the test set to obtain test accuracy, and obtaining the scoring model under the condition that the test accuracy is greater than or equal to a preset accuracy threshold.

It should be noted that, in the embodiment of the invention, by combining the big data technology and the AI algorithm, the scoring model is constructed, the scoring model is used for comprehensively scoring all commodities in the mall, the construction of the scoring model firstly needs to collect the commodity information of each commodity in the mall, including the characteristic identifier of each commodity and the characteristic scoring value in the historical time period of the commodity, the sample data can be constructed according to the collected data, the sample data is split into a training set and a testing set, then the training set is adopted to train the classifier to obtain an initial scoring model, then the initial scoring model is tested based on the testing set, the trained scoring model is obtained after the testing passes, the scoring model can be used for outputting the scoring values of all commodity characteristics in the mall at regular time to obtain a commodity scoring table, or the commodity identifier or the commodity characteristic identifier can be input into the scoring model, meanwhile, the scoring model needs to be updated according to the characteristic index, and the sample data used by the model is ensured to be the latest.

It should be noted that, the invention adopts the mode of stacking and cross-verifying classification algorithm to carry out model training, and outputs commodity scoring table, the model training algorithm comprises SVM, GBDT, random forest and XGBOOST, the SVM separates different types of samples into hyperplanes in sample space, and the optimal hyperplanes are searched by replacing nonlinear mapping to high-dimensional space with inner product kernel function, thus achieving the purposes of optimal classification and prediction; GBDT is also called gradient lifting decision tree, GBDT uses CART regression tree as classifier, generates a weak classifier through multiple iterations, determines the parameters of the next weak classifier through experience risk minimization, trains each classifier on the basis of the residual error of the previous classifier, continuously fits square loss function, and finally obtains strong classifier through linear superposition of weak learners of multiple iterations; the random forest algorithm establishes a forest model in a random sampling and average voting mode, the random forest comprises a plurality of independent decision trees, and classification prediction is finally made through the fact that the mode of the doctor is long, the minority is subject to majority; the XGBOOST classification algorithm and the GBDT belong to Boosting integrated learning, the output result of each weak learner is regarded as a continuous value, each iteration is developed based on the Taylor second derivative of the loss function of the previous iteration, and the square loss function is fitted by continuously optimizing the negative gradient and the residual error, so that the performance and the accuracy of the model are finally improved.

In the embodiment of the invention, SVM, GBDT and random forest are used as a base learner, XGBOOST is used as a weight classification learner, the result predicted by the base learner is poured into XGBOOST, the same weight is given at the beginning, the parameter entering is continuously optimized along with training, proper weight parameters are allocated, and finally the result output of XGBOOST is used as a model scoring result.

Optionally, the step of constructing the sample data based on the commodity information includes: preprocessing commodity information, wherein the preprocessing at least comprises the following steps: desensitizing treatment, data cleaning treatment and data missing value treatment; and converting the preprocessed commodity information into vector data to obtain a commodity vector data set, and constructing sample data based on the commodity vector data set.

Note that, when the scoring model is constructed, the obtained features are presented in the form of characters, in order to construct the scoring model, the feature data needs to be converted into vector data used by machine learning, and before vector conversion is performed, the original data needs to be preprocessed, including desensitizing sensitive data, cleaning data and processing data missing values, when the data missing values are processed, a mean/median/mode interpolation method, a fixed value substitution method, a nearest neighbor interpolation method, a regression method, an interpolation method and the like can be adopted, and meanwhile, for data with anomalies, the following methods can be adopted to eliminate anomalies: firstly, business rule intervention is performed, and data which obviously does not accord with business logic are filtered and discarded; secondly, according to whether the upper and lower bounds, the upper and lower quartiles, the median and the mean of the data accord with normal distribution, unreasonable data are avoided through a data mining method.

Step S101, responding to a commodity searching request of a user side to obtain a user identification of the user side.

When a user logs in the platform, browsing behaviors of the user are collected in real time, user identification of the user side is obtained according to a search request of the user, and required commodities corresponding to the user are determined according to the search request of the user.

Step S102, a historical behavior data table of a user side is obtained based on a user identifier, and a commodity scoring table and a commodity dictionary table are obtained, wherein the historical behavior data table at least comprises: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics.

Optionally, the step of acquiring the historical behavior data table of the user terminal based on the user identifier includes: collecting shopping behavior data of a user terminal based on a user identifier, wherein the shopping behavior data at least comprises: commodity identification and behavior characteristics; configuring a label value for each commodity based on the behavior characteristics to obtain a commodity label value; and generating a historical behavior data table of the user terminal based on the user identification, the commodity identification and the commodity label value.

After responding to the commodity searching request of the user side, firstly, acquiring a historical behavior data table corresponding to the user identifier of the user side, wherein the historical behavior data is established according to the historical shopping behavior data of the user side, firstly, acquiring shopping behaviors of the user on all commodities, configuring a label value for the commodities corresponding to the behavior characteristics to obtain the commodity label value, and then generating the historical behavior data table according to the user identifier, the commodity identifier and the commodity label value, and the behavior characteristics comprise: the method comprises the steps of commodity evaluation, commodity forwarding recommendation, commodity collection, commodity purchasing, text comment, blueprint comment, commodity ordering, commodity browsing, commodity homepage stay time, skip page number and the like, wherein when labels are configured, label values of 1-5 can be configured for the commodities according to the star level of the evaluation for commodity evaluation, and label configurations of 0 and 1 can be configured for the commodities according to whether the commodity forwarding recommendation behavior is recommended and forwarded.

Optionally, the step of obtaining the commodity dictionary table includes: information acquisition is carried out on each commodity in the mall to obtain basic commodity information, wherein the basic commodity information at least comprises: commodity identification and commodity characteristics; and assigning a weight value to each commodity feature to obtain feature weights, and generating a commodity dictionary table based on the commodity identifications, the commodity features and the feature weights.

The commodity dictionary table reflects the basic weight of each commodity basically, the weight information is derived from commodity brands, commodity versions, merchant strength, credit, off-line channel layout and the like, and is used for objectively reflecting initial value and initial score of the commodity, so that the problem of cold start in a recommendation system can be well solved, and comprehensive evaluation is carried out on the commodity more comprehensively.

Meanwhile, in order to cope with the problem of recommending cold start, in data acquisition, besides paying attention to business behavior data, the acquisition of commodity and user basic data can be paid attention to, for example, user basic information such as age, gender, occupation, region and the like is acquired, and the data can be used as a coarse-granularity personalized recommendation basis for users; acquiring expert knowledge, and rapidly establishing an initial scoring table (corresponding to the commodity dictionary table) of the commodity by using authoritative experts, deep shoppers, marketers and marketers for evaluating the commodity; basic information of commodities, such as manufacturer, off-line public praise, brand awareness, content popularization information, production date and the like, is collected and used as a basis for rapidly establishing an initial commodity scoring table (corresponding to the commodity dictionary table).

And step S103, constructing a commodity correlation characteristic table based on the historical behavior data table and the commodity scoring table, and constructing a commodity characteristic weight table based on the historical behavior data table and the commodity dictionary table.

It should be noted that, according to the three basic tables including the historical behavior data table, the commodity scoring table and the commodity dictionary table, two sub-tables may be generated, which are the commodity correlation characteristic table and the commodity characteristic weight table respectively.

Optionally, the step of constructing the merchandise association feature table based on the historical behavior data table and the merchandise scoring table includes: combining each commodity in the commodity scoring table with each commodity in the historical behavior data table to obtain a set of associated commodity pairs; calculating a weight value of each associated commodity pair in the associated commodity pair set; calculating the association weight value of each commodity in the commodity scoring table and the user side based on the weight value of the associated commodity pair; and generating a commodity association characteristic table based on the association weight value of each commodity and the user side.

It should be noted that, the commodity association feature table is used for associating the new commodity which is not browsed or purchased by the user with the commodity which is browsed or purchased by the user, calculating the association degree between the new commodity and the user, and further determining whether the new commodity needs to be recommended to the user.

When the commodities are associated, combining each commodity in the commodity grading table with each commodity in the historical behavior data table to obtain an associated commodity pair, and then calculating the weight value of the associated commodity pair according to the weight value of each associated commodity in the commodity pair, and further calculating the weight value between each commodity and the user side recorded in the commodity grading table to obtain the associated feature table.

Optionally, after generating the commodity recommendation list based on the commodity list, further comprising: determining commodities to be recommended based on the commodity recommendation list; arranging commodities to be recommended, rendering a terminal page of a user terminal based on the commodities to be recommended, and displaying the commodities to be recommended to the terminal page of the user terminal.

After two sub-tables are generated according to the historical behavior data table, the commodity evaluation table and the commodity dictionary table, each sub-table records a weight value corresponding to the user side of each commodity or a weight value corresponding to the user side of commodity characteristics, the weight value is calculated according to the commodity weight values recorded in the table to obtain a comprehensive weight value of each commodity in the upper mall, the weight values are ordered, N commodities with the weight values arranged in front are selected, and a commodity recommendation list is generated according to the selected commodities.

After generating the commodity recommendation list of the user side, the commodity recommendation system interacts with the terminal interface of the user side, arranges the recommended commodities according to labels such as guessing that the user likes, personalized recommendation and the like, and displays the commodities interested by the user side in the mall on the terminal interface of the user side based on the arranged commodities.

The following detailed description is directed to alternative embodiments.

The commodity recommendation system is divided into four layers of structures, namely a data source layer, a data warehouse layer, a computing system and an application layer.

The data source layer is used as a system bottom layer and is used for collecting user historical data and commodity basic information, user real-time behavior data such as browsing behavior data, purchasing behavior data, clicking behavior data and the like, the user real-time behavior data can flow into an upper layer through a front-end buried point, kafka and a link pipeline, the real-time performance of data flow is achieved, shopping behavior and commodity preference of a user in a certain time segment of an e-commerce platform are reflected in real time, the user historical business data are business data stored in a data warehouse based on a historical time period, the data are stored in an HDFS database, the data volume is large and are not easy to lose, the recommendation system calculates, analyzes and operates based on data sources collected by the data source layer, and related recommendation preference content is produced.

The middle two-layer computing system and the data warehouse layer belong to an infrastructure layer, and the data warehouse layer can execute the most basic data computation, so as to complete the basic functions of user portraits, ETL data cleaning and filtering, model training set construction, basic data analysis, index list construction and the like, and provide high-quality training sets and analysis references for the upper-layer computing system.

The computing system follows the three-section recommendation strategy of recall, sorting and rearrangement of the recommendation procedure, which covers the machine learning and deep learning algorithm model, and aims to produce the model file (corresponding to the commodity scoring table) of the recommendation system, wherein the model file can be understood to be an algorithm scoring card, and the personalized recommendation of commodities to users is sequentially sorted based on the algorithm scoring card. Specifically, the recall model is divided into hot recall, collaborative filtering, commodity preference and association rules, and the main purpose is to coarsely screen a user basic characteristic, a behavior characteristic and a massive commodity library through recall, filter and delete unnecessary data from massive data, lighten the burden for subsequent data training, improve the training efficiency and training quality and improve the real-time performance of recommended services. The sorting is to sort the training set and score the commodity by using machine learning and deep learning methods, and output the training set and commodity in the form of model files. The rearrangement is to implement manual intervention, operate and regulate, surround user experience, perform proper de-duplication, break up, enrich diversity, raise commodity weight, lower commodity weight, and set top and bottom, so as to realize flexible commodity recommendation distribution and improve user experience.

The application layer is directly user-oriented, the recommendation list and the commodity details of the lower layer are cached in a Redis database and displayed on a terminal page of a user side, and the content can be: guessing that you like, personalizing the recommendation, retail goods, etc., puts in the recommendation stream to the user, forming a good user experience.

FIG. 2 is a schematic diagram of an alternative commodity recommendation process according to an embodiment of the present invention, where the commodity recommendation process includes two parts, namely online recommendation and offline training, as shown in FIG. 2, after a user logs into a shopping platform, a commodity recommendation system of the shopping platform performs recommendation service to the user, the online recommendation includes four parts, namely service routing, feature indexing, feature extraction and prediction scoring, the offline training includes five parts, namely sample extraction, feature extraction, training set generation, feature index generation and training model, the offline training obtains sample data from a database, stores a constructed prediction model into REDIS cache, and directly invokes the model to evaluate a commodity when the commodity recommendation service is required.

FIG. 3 is a flowchart of an alternative online ordering of commodities according to an embodiment of the present invention, as shown in FIG. 3, three data tables including a commodity scoring table, a historical behavior data table and a commodity dictionary table are first established, features and weight values are stored in each table, then a commodity association feature table and a commodity basic weight table are constructed based on the commodity scoring table and the historical behavior data table, the two tables are associated through user identification, then weight ordering calculation is performed on the commodity association feature table and the commodity basic weight table, a commodity recommendation list is obtained, and finally the commodity in the commodity recommendation list is displayed to a terminal interface of a user side by a commodity recommendation system.

The following describes in detail another embodiment.

Example two

The commodity recommendation apparatus provided in this embodiment includes a plurality of implementation units, each of which corresponds to each implementation step in the above-described embodiment.

FIG. 4 is a schematic diagram of an alternative merchandise recommendation apparatus according to an embodiment of the invention, as shown in FIG. 4, the merchandise recommendation apparatus comprising: a response unit 41, an acquisition unit 42, a construction unit 43, a generation unit 44, wherein,

a response unit 41, configured to respond to a commodity search request from a user side, and obtain a user identifier of the user side;

the obtaining unit 42 is configured to obtain a historical behavior data table of the user side based on the user identifier, and obtain a commodity scoring table and a commodity dictionary table, where the historical behavior data table at least includes: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics;

a construction unit 43 for constructing a commodity correlation feature table based on the historical behavior data table and the commodity scoring table, and constructing a commodity feature weight table based on the historical behavior data table and the commodity dictionary table;

The generating unit 44 is configured to calculate and sort weight values of the commodities in the commodity correlation feature table and the commodity feature weight table, obtain a commodity list, and generate a commodity recommendation list based on the commodity list.

The commodity recommending device responds to the commodity searching request of the user side through the responding unit 41 to obtain the user identification of the user side; the obtaining unit 42 obtains a historical behavior data table of the user terminal based on the user identifier, and obtains a commodity scoring table and a commodity dictionary table, where the historical behavior data table at least includes: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics; building a commodity correlation feature table based on the historical behavior data table and the commodity scoring table by a building unit 43, and building a commodity feature weight table based on the historical behavior data table and the commodity dictionary table; the commodity correlation characteristic table and the commodity in the commodity characteristic weight table are subjected to weight value calculation and sorting by the generating unit 44 to obtain a commodity list, and a commodity recommendation list is generated based on the commodity list.

Optionally, the acquiring unit includes: the first acquisition module is used for acquiring shopping behavior data of the user terminal based on the user identifier, wherein the shopping behavior data at least comprises: commodity identification and behavior characteristics; the first configuration module is used for configuring label values for each commodity based on the behavior characteristics to obtain commodity label values; the first generation module is used for generating a historical behavior data table of the user side based on the user identification, the commodity identification and the commodity label value.

Optionally, the acquisition unit further includes: the second acquisition module is used for acquiring information of each commodity in the mall to obtain basic commodity information, wherein the basic commodity information at least comprises: commodity identification and commodity characteristics; the first giving module is used for giving a weight value to each commodity feature to obtain a feature weight, and generating a commodity dictionary table based on commodity identifications, commodity features and the feature weight.

Optionally, the building unit comprises: the first combination module is used for combining each commodity in the commodity scoring table with each commodity in the historical behavior data table to obtain a related commodity pair set; the first calculation module is used for calculating the weight value of each associated commodity pair in the associated commodity pair set; the first calculation module is used for calculating the association weight value of each commodity in the commodity scoring table and the user side based on the weight value of the associated commodity pair; and the second generation module is used for generating a commodity association characteristic table based on the association weight value of each commodity and the user side.

Optionally, the commodity recommendation device further includes: the third acquisition module is used for acquiring commodity information of each commodity in the mall, wherein the commodity information at least comprises: commodity feature identification and commodity feature scoring values; the first construction module is used for constructing sample data based on commodity information and dividing the sample data into a training set and a testing set based on a preset proportion; the first training module is used for training the classifier based on the training set to obtain an initial scoring model; the first test module is used for testing the initial scoring model based on the test set to obtain test accuracy, and the scoring model is obtained under the condition that the test accuracy is greater than or equal to a preset accuracy threshold.

Optionally, the first building module includes: the first preprocessing sub-module is used for preprocessing commodity information, wherein the preprocessing at least comprises: desensitizing treatment, data cleaning treatment and data missing value treatment; the first conversion sub-module is used for converting the preprocessed commodity information into vector data to obtain a commodity vector data set, and constructing sample data based on the commodity vector data set.

The commodity recommending apparatus may further include a processor and a memory, wherein the response unit 41, the acquisition unit 42, the construction unit 43, the generation unit 44, and the like are stored in the memory as program units, and the processor executes the program units stored in the memory to realize the corresponding functions.

The processor includes a kernel, and the kernel fetches a corresponding program unit from the memory. The kernel can be provided with one or more than one, and the commodity is recommended to the user side by adjusting the kernel parameters.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), which includes at least one memory chip.

According to another aspect of the embodiment of the present application, there is also provided a computer readable storage medium, where the computer readable storage medium includes a stored computer program, and when the computer program is executed, the apparatus on which the computer readable storage medium is located is controlled to execute any one of the commodity recommendation methods described above.

According to another aspect of the embodiments of the present application, there is also provided an electronic device including one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement any one of the commodity recommendation methods described above.

The application also provides a computer program product adapted to perform, when executed on a data processing device, a program initialized with the method steps of: responding to a commodity searching request of a user side to obtain a user identification of the user side; acquiring a historical behavior data table of a user terminal based on a user identifier, and acquiring a commodity scoring table and a commodity dictionary table, wherein the historical behavior data table at least comprises: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics; building a commodity correlation characteristic table based on the historical behavior data table and the commodity scoring table, and building a commodity characteristic weight table based on the historical behavior data table and the commodity dictionary table; and calculating and sequencing weight values of the commodities in the commodity association characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list.

Fig. 5 is a block diagram of a hardware structure of an electronic device (or mobile device) of a commodity recommendation method according to an embodiment of the present invention. As shown in fig. 5, the electronic device may include one or more (shown in fig. 5 as 502a, 502b, … …,502 n) processors 502 (the processors 502 may include, but are not limited to, a microprocessor MCU, a programmable logic device FPGA, etc. processing means), a memory 504 for storing data. In addition, the method may further include: a display, an input/output interface (I/O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the I/O interface), a network interface, a keyboard, a power supply, and/or a camera. It will be appreciated by those of ordinary skill in the art that the configuration shown in fig. 5 is merely illustrative and is not intended to limit the configuration of the electronic device described above. For example, the electronic device may also include more or fewer components than shown in FIG. 5, or have a different configuration than shown in FIG. 5.

The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.

In the foregoing embodiments of the present invention, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.

In the several embodiments provided in the present application, it should be understood that the disclosed technology may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and the division of the units, for example, may be a logic function division, and may be implemented in another manner, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing program codes.

The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.

Claims

1. A commodity recommendation method, comprising:

responding to a commodity searching request of a user side to obtain a user identification of the user side;

acquiring a historical behavior data table of the user terminal based on the user identifier, and acquiring a commodity scoring table and a commodity dictionary table, wherein the historical behavior data table at least comprises: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: the commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics;

constructing a commodity correlation feature table based on the historical behavior data table and the commodity scoring table, and constructing a commodity feature weight table based on the historical behavior data table and the commodity dictionary table;

and calculating and sequencing weight values of the commodities in the commodity association characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list.

2. The commodity recommendation method according to claim 1, further comprising, after generating a commodity recommendation list based on the commodity list:

determining commodities to be recommended based on the commodity recommendation list;

Arranging the commodities to be recommended, rendering the terminal page of the user terminal based on the commodities to be recommended, and displaying the commodities to be recommended to the terminal page of the user terminal.

3. The merchandise recommendation method according to claim 1, wherein the step of obtaining the historical behavior data table of the user side based on the user identification comprises:

collecting shopping behavior data of the user terminal based on the user identifier, wherein the shopping behavior data at least comprises: the commodity identification and behavior characteristics;

configuring a label value for each commodity based on the behavior characteristics to obtain a commodity label value;

and generating the historical behavior data table of the user terminal based on the user identification, the commodity identification and the commodity label value.

4. The commodity recommendation method according to claim 1, wherein the step of obtaining a commodity dictionary table comprises:

information acquisition is carried out on each commodity in the mall to obtain basic commodity information, wherein the basic commodity information at least comprises: commodity identification and commodity characteristics;

and assigning a weight value to each commodity feature to obtain the feature weight, and generating the commodity dictionary table based on the commodity identification, the commodity feature and the feature weight.

5. The commodity recommendation method according to claim 1, wherein the step of constructing a commodity correlation feature table based on the historical behavior data table and the commodity scoring table comprises:

combining each commodity in the commodity scoring table with each commodity in the historical behavior data table to obtain a set of associated commodity pairs;

calculating a weight value of each associated commodity pair in the associated commodity pair set;

calculating the association weight value of each commodity in the commodity scoring table and the user side based on the weight value of the associated commodity pair;

and generating the commodity association characteristic table based on the association weight value of each commodity and the user side.

6. The merchandise recommendation method according to claim 1, wherein the scoring model is pre-constructed, and the step of constructing the scoring model comprises:

acquiring commodity information of each commodity in a mall, wherein the commodity information at least comprises: commodity feature identification and commodity feature scoring values;

constructing sample data based on the commodity information, and dividing the sample data into a training set and a testing set based on a preset proportion;

training the classifier based on the training set to obtain an initial scoring model;

And testing the initial scoring model based on the test set to obtain test accuracy, and obtaining the scoring model under the condition that the test accuracy is greater than or equal to a preset accuracy threshold.

7. The commodity recommendation method according to claim 6, wherein the step of constructing sample data based on the commodity information includes:

preprocessing the commodity information, wherein the preprocessing at least comprises the following steps: desensitizing treatment, data cleaning treatment and data missing value treatment;

and converting the preprocessed commodity information into vector data to obtain a commodity vector data set, and constructing the sample data based on the commodity vector data set.

8. A commodity recommendation device, comprising:

the response unit is used for responding to the commodity searching request of the user side and obtaining the user identification of the user side;

the acquiring unit is configured to acquire a historical behavior data table of the user terminal based on the user identifier, and acquire a commodity scoring table and a commodity dictionary table, where the historical behavior data table at least includes: the commodity identification and the commodity label value are obtained based on a grading model, and the commodity dictionary table at least comprises: the commodity identification, commodity characteristics and characteristic weights corresponding to the commodity characteristics;

The construction unit is used for constructing a commodity correlation characteristic table based on the historical behavior data table and the commodity scoring table, and constructing a commodity characteristic weight table based on the historical behavior data table and the commodity dictionary table;

and the generation unit is used for calculating and sequencing the weight values of the commodities in the commodity association characteristic table and the commodity characteristic weight table to obtain a commodity list, and generating a commodity recommendation list based on the commodity list.

9. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored computer program, wherein the computer program, when run, controls a device in which the computer readable storage medium is located to perform the commodity recommendation method according to any one of claims 1 to 7.

10. An electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the merchandise recommendation method of any one of claims 1 to 7.