Target user implicit relationship classification method based on multiple semantic factors and feature aggregation

Abstract

The invention discloses a target user implicit relationship classification method based on multiple semantic factors and feature aggregation. The method comprises the following steps: (1) carrying out three types of local semantic feature extraction on event texts of known target users; (2) carrying out global semantic feature extraction on the event texts of the known target users; (3) carrying out pooling processing and weighted fusion on three types of local semantic features of the event texts, and inputting the processed local semantic features into a self-attention network to obtain multi-semantic factor aggregation feature vectors of the texts; and (4) inputting multi-semantic factor aggregation features and global semantic features into a trained classifier, and obtaining a relationship category between the target users after performing softmax on output features. According to the target user implicit relationship classification method based on multiple semantic factors and feature aggregation, implicit relationships between users involved in an electronic commerce activity can be effectively mined, and the cognition efficiency and analysis efficiency of a recommendation system on relationships between target users are improved.

Description

technical field [0001] The invention belongs to the technical field of data mining, and more specifically relates to a method for classifying implicit relationships of target users based on multi-semantic factors and feature aggregation. Background technique [0002] There are a large number of users on the e-commerce platform, and these users have diverse attributes and behaviors, and there are various connections between users that cannot be directly obtained through explicit information, so as to mine the implicit information between users who use the e-commerce platform. Relationships are increasingly becoming an important requirement in the field of personalized recommendation. With the increasing number and complexity of e-commerce activities, for the purpose of analyzing user groups and optimizing recommendations, various researches on the analysis of target users involved in e-commerce are becoming more and more extensive. One of the mainstream methods is to analyze ...

AI Technical Summary

Problems solved by technology

The traditional pattern matching method to extract the relationship between target users depends on the established rules and initial seeds, and there will be problems such as data sparsity, while the machine learning method relies on the size and quality of the manually labeled target user background knowledge data set and manual feature design. The rationality, the effect is not good

Images