CN-122022937-A - Online commodity recommendation method and system based on user behavior analysis

CN122022937ACN 122022937 ACN122022937 ACN 122022937ACN-122022937-A

Abstract

The invention discloses an online commodity recommendation method and system based on user behavior analysis, and relates to the technical field of data processing, wherein the method comprises the steps of reading a plurality of recommendation requests of a target user, collecting a plurality of commodity recommendation sequences and a plurality of behavior data records; the method comprises the steps of establishing a request constraint structure, analyzing and identifying forward trend evolution characteristics and reverse trend evolution characteristics, establishing forward filtering characteristics and reverse filtering characteristics, constructing a recommended commodity library through commodity filtering, identifying popularity and exposure of each commodity in the recommended commodity library, executing popularity and exposure weight reduction based on commodity interaction characteristics, and generating a recommended commodity sequence. The technical problems that the commodity recommendation deviation is caused by incapability of effectively capturing and responding to the dynamic user intention evolution and excessive exposure of the popular commodity in the prior art, and further the commodity recommendation real-time responsiveness and dynamic adaptability are insufficient are solved, and the technical effect of improving the commodity recommendation real-time responsiveness and dynamic adaptability is achieved.

Inventors

ZHAO YONGJIA
YE MINMIN
LIN WEIYI
YE KUN
Li feida
GAO SIFAN
CHEN YITAI

Assignees

浙建云采(龙游)科技有限责任公司

Dates

Publication Date: 20260512
Application Date: 20251205

Claims (10)

1. The online commodity recommending method based on the user behavior analysis is characterized by comprising the following steps of: Reading a plurality of recommendation requests of a target user with similar commodity attributes in a preset time zone, and collecting a plurality of commodity recommendation sequences and a plurality of behavior data records; Performing constraint increment identification of a request field on the plurality of recommendation requests, establishing a request constraint structure, analyzing the multi-round behavior data record and the multi-round commodity recommendation sequence, and identifying forward trend evolution characteristics and reverse trend evolution characteristics corresponding to the request constraint structure; Establishing forward filtering characteristics and reverse filtering characteristics according to the forward trend evolution characteristics and the reverse trend evolution characteristics, and constructing a recommended commodity library through commodity filtering; And identifying popularity and exposure degree of each commodity in the recommended commodity library, executing popularity and exposure degree weight reduction based on commodity interaction characteristics, and generating a recommended commodity sequence.
2. The online commodity recommending method based on user behavior analysis according to claim 1, wherein the steps of reading a plurality of recommending requests of a target user with similar commodity attributes in a preset time zone, collecting a plurality of commodity recommending sequences and a plurality of behavior data records, and comprise: Reading all recommendation requests of a target user in a preset time zone; Extracting request fields corresponding to all recommendation requests respectively, performing part-of-speech analysis to identify noun fields, and classifying the recommendation requests with the same noun fields into a plurality of recommendation requests with similar commodity attributes; Acquiring recommended commodity sequencing features generated by an online commodity platform in response to the plurality of recommendation requests, and generating the multi-round commodity recommendation sequence; And acquiring interesting behaviors of the target user based on the multi-round commodity recommendation sequence, and generating the multi-round behavior data record.
3. The online commodity recommendation method based on user behavior analysis according to claim 1, wherein performing constraint increment identification of a request field on the plurality of recommendation requests, establishing a request constraint structure, comprises: extracting a plurality of request fields of the plurality of recommendation requests; And carrying out hierarchical connection and labeling according to the change states of the request fields among the plurality of request fields, and establishing the request constraint structure, wherein the request constraint structure is represented by a graph structure and comprises a plurality of field evolution paths.
4. The online commodity recommending method based on user behavior analysis according to claim 3, wherein establishing the request constraint structure by performing hierarchical connection and labeling according to a request field change state between the plurality of request fields comprises: Determining a root request in the plurality of recommendation requests, identifying request field change information of a first recommendation request from the root request to the rest recommendation requests, performing hierarchical connection and labeling, and establishing a primary constraint structure; Extracting a second recommendation request from the remaining recommendation requests, identifying request field change information of the root request and the first recommendation request, establishing a field evolution relation of the second recommendation request, and performing hierarchical annotation optimization on the primary constraint structure to generate a secondary constraint structure; And continuously extracting a third recommendation request from the remaining recommendation requests, analyzing field evolution relations with the root request, the first recommendation request and the second recommendation request, and performing hierarchical annotation optimization until each request in the remaining recommendation requests is traversed, so as to generate the request constraint structure.
5. The online commodity recommendation method based on user behavior analysis according to claim 4, wherein analyzing the multiple rounds of behavior data records and the multiple rounds of commodity recommendation sequences identifies forward trend evolution features and reverse trend evolution features corresponding to the request constraint structure comprises: Analyzing the user interest scoring information under each field evolution path node in the request constraint structure by a preset interest scoring mechanism according to the multi-round behavior data record, and determining a scoring rising path and a scoring falling path; Constructing the forward trend evolution characteristic according to paths with scores higher than a first preset threshold value in paths with scores rising; And constructing the reverse trend evolution characteristic according to paths with scores lower than a second preset threshold value in paths with scores falling.
6. The online commodity recommendation method based on user behavior analysis according to claim 1, wherein establishing a forward filtering feature and a reverse filtering feature with the forward trend evolution feature and the reverse trend evolution feature, establishing a recommended commodity library through commodity filtering, comprises: establishing the forward filtering feature with the forward trend evolution feature; establishing the reverse filtering feature with the reverse trend evolution feature; and screening a primary screening commodity library with the matching degree of the primary screening commodity library and the forward filtering characteristic being larger than a preset matching threshold value from the original commodity library, screening commodities with the matching degree of the reverse filtering characteristic being larger than the preset matching threshold value from the primary screening commodity library by using the reverse filtering characteristic, deleting the commodities from the primary screening commodity library, and generating the recommended commodity library.
7. The online commodity recommending method based on user behavior analysis according to claim 1, wherein identifying popularity and exposure of each commodity in the recommended commodity library, performing popularity and exposure downweighting based on commodity interaction characteristics, and generating a recommended commodity sequence comprises: collecting historical multi-user interaction records of each commodity, and constructing a commodity interaction diagram; Identifying popularity and exposure of each commodity based on the commodity interaction diagram; Analyzing the similarity between each commodity feature and forward filtering feature in the recommended commodity library, and establishing original recommended weight distribution; Calculating the weight reduction coefficient of each commodity by taking the popularity and the exposure of each commodity as input characteristics, and correcting the original recommended weight distribution to generate a depolarization weight matrix; and using the unbiased weight matrix for recommendation management of the recommended commodity library to generate the recommended commodity sequence.
8. The online commodity recommendation method according to claim 7, wherein popularity is determined by analyzing a transaction completion number of each commodity in the commodity interaction diagram, and exposure is determined by analyzing a total number of times each commodity in the commodity interaction diagram is presented to all users.
9. The online commodity recommending method based on the user behavior analysis according to claim 7, wherein calculating the weight-reducing coefficient of each commodity with popularity and exposure of each commodity as input features comprises: Collecting historical online transaction data of the target user, and analyzing the correlation between the transaction completion degree and the exposure degree and popularity; The weight reduction coefficient is generated based on popularity and exposure of the correlation to each commodity.
10. An online merchandise recommendation system based on user behavior analysis, wherein the system is for implementing the online merchandise recommendation method based on user behavior analysis according to any one of claims 1 to 9, the system comprising: The data acquisition module is used for reading a plurality of recommendation requests of the target user with similar commodity attributes in a preset time zone, and acquiring a plurality of commodity recommendation sequences and a plurality of behavior data records; the trend evolution feature recognition module is used for carrying out constraint increment recognition of request fields on the plurality of recommendation requests, establishing a request constraint structure, analyzing the multi-round behavior data record and the multi-round commodity recommendation sequence, and recognizing forward trend evolution features and reverse trend evolution features corresponding to the request constraint structure; the recommended commodity library construction module is used for building forward filtering characteristics and reverse filtering characteristics according to the forward trend evolution characteristics and the reverse trend evolution characteristics, and constructing a recommended commodity library through commodity filtering; And the recommended commodity sequence generation module is used for identifying popularity and exposure of each commodity in the recommended commodity library, executing popularity and exposure weight reduction based on commodity interaction characteristics and generating a recommended commodity sequence.

Description

Online commodity recommendation method and system based on user behavior analysis Technical Field The application relates to the technical field of data processing, in particular to an online commodity recommendation method and system based on user behavior analysis. Background With the rapid development of electronic commerce and online service platforms, commodity recommendation systems become the key for improving user experience and promoting consumption decisions, traditional commodity recommendation mainly relies on collaborative filtering, content filtering or mixed recommendation models, a personalized recommendation list is generated by analyzing historical behaviors and commodity attributes of users, however, certain limitations are presented when the users face dynamic and changeable preferences, on one hand, user interests can be remarkably evolved over time or situation changes, on the other hand, frequent recommendation requests and implicit deep intention in short-term behavior sequences are difficult to effectively capture, and recommendation results deviate from real demands of the users. In an actual application scene, a user often initiates a plurality of recommendation requests for similar commodities in a short time, for example, the user may query similar products of different brands for a plurality of times or repeatedly adjust screening conditions, the existing method lacks continuous analysis on trend evolution in a plurality of rounds of interaction sequences, and in addition, recommendation deviation is easily caused by commodity popularity and exposure, namely, popular commodities or over-exposed commodities may mask long-tail commodities really interested by the user, so that the diversity and novelty of recommendation are reduced. Therefore, in the related technology at the present stage, the technical problems that the real-time response and the dynamic adaptability of commodity recommendation are insufficient due to commodity recommendation deviation caused by incapability of effectively capturing and responding to the evolution of the dynamic user intention and the overexposure of hot commodities exist. Disclosure of Invention The online commodity recommendation method and system based on user behavior analysis solve the technical problems that commodity recommendation deviation is caused by incapability of effectively capturing and responding to dynamic user intention evolution and excessive exposure of hot commodities in the prior art, and further the commodity recommendation is insufficient in real-time responsiveness and dynamic adaptability, and achieve the technical effect of improving the real-time responsiveness and the dynamic adaptability of commodity recommendation. The application provides an online commodity recommending method based on user behavior analysis, which comprises the steps of reading a plurality of recommending requests of a target user with similar commodity attributes in a preset time zone, collecting a plurality of commodity recommending sequences and a plurality of behavior data records, carrying out constraint increment recognition of request fields on the plurality of recommending requests, establishing a request constraint structure, analyzing the plurality of behavior data records and the plurality of commodity recommending sequences, recognizing forward trend evolving characteristics and reverse trend evolving characteristics corresponding to the request constraint structure, establishing forward trend evolving characteristics and reverse trend evolving characteristics according to the forward trend evolving characteristics and the reverse trend evolving characteristics, constructing a recommending commodity library through commodity filtering, recognizing popularity and exposure of each commodity in the recommending commodity library, and executing popularity and exposure weight reduction based on commodity interaction characteristics to generate the recommending commodity sequence. In a possible implementation manner, the online commodity recommending method based on the user behavior analysis further performs the following processing of reading all recommending requests of a target user in a preset time zone, extracting request fields corresponding to all recommending requests respectively, performing part-of-speech analysis to identify noun fields, classifying the recommending requests with the same noun fields into the plurality of recommending requests with similar commodity attributes, collecting recommending commodity ordering characteristics generated by an online commodity platform in response to the plurality of recommending requests, generating a multi-round commodity recommending sequence, collecting interesting behaviors of the target user based on the multi-round commodity recommending sequence, and generating the multi-round behavior data record. In a possible implementation manner, the online commodity recommending method b