Score recommendation method based on heterogeneous information

Abstract

The invention provides a score recommendation method based on heterogeneous information. The method includes the following steps: 1, obtaining user ID, article ID, score information, comment information and article description information of an article; 2, converting the score information, the comment information and the article description information into a score vector, a comment vector and an article description vector respectively; 3, introducing the score vector, the comment vector and the article description vector into a loss function, and solving is conducted by the adoption of a gradient descent method; 4, according to final vectors obtained through calculation in the step 3, a user recommendation degree of the article is obtained through equations. By the adoption of the score recommendation method, the short text problem of text messages is solved, different information under different space expression can be integrated more effectively, so that the score information, the comment information and the article information are fused better, and finally, more accurate recommendation is made.

Description

technical field [0001] The invention relates to the field of data mining, and relates to a rating recommendation method based on heterogeneous information. Background technique [0002] With the rapid development of information technology, various technologies of big data have emerged as the times require, but the problem of information overload is still becoming more and more serious. As an important tool to solve information overload, the rating recommendation system recommends more accurate items to users by analyzing the behavior information between users and items, and the description information of users and items, so as to maximize the value of information. [0003] In rating recommendation systems, the challenge is mainly the sparsity of data. In addition to rating information, text information such as comments has become the best supplement. However, in the research on integrating rating and review information into rating prediction recommendation system, most of ...

AI Technical Summary

Problems solved by technology

However, such methods suffer from the fact that text messages are mostly short text

Moreover, the inaccuracy of text information also poses certain challenges to traditional methods

Images