Weighted averaging personalized friend recommendation method based on LDA (Linear Discriminant Analysis)

Abstract

The invention discloses a weighted averaging personalized friend recommendation method based on LDA (Linear Discriminant Analysis). The method comprise the following contents: obtaining the first-degree, second-degree,..., m-degree friends of a target user t in a social network to form a candidate friend set C, and extracting the target user and the static attribute information of each user in theset C; utilizing an LDA topic modeling method, analyzing a topic to which the user pays attention so as to deduce the hobbies, the interests, the status and the age and gender information of the candidate user; analyzing the dynamic behavior information in the social network, and using a weighted average method to calculate a similarity among users; and finally, obtaining the grading vector of the target user for the candidate users, carrying out sorting according to the grading vector, and recommending the candidate users which rank in the Top-N to the target user. By use of the method, theattribute information of the target user and the candidate users can be more comprehensively considered so as to improve friend recommendation accuracy to a maximum degree. By use of the method, verification is carried out on a microblog dataset, and an experiment result indicates that the method has a good recommendation effect.

Description

technical field [0001] The invention belongs to the technical field of the Internet, relates to social network recommendation under the Internet, in particular to a method for recommending personalized friends based on LDA weighted average. Background technique [0002] With the vigorous development of Web2.0 technology, the world has gradually ushered in the era of social network (SocialNetwork). Some representative social networking sites have become influential information platforms, such as: Facebook, Twitter and Sina Weibo. They combine user groups and information together, so that users can share and obtain information conveniently, and also greatly expand the user's social groups. However, with the rapid increase of social network users, the amount of information in the social network increases sharply. For social network users, how to find suitable friends in these huge data and expand their social network friend circle has become a difficult problem. [0003] In or...

AI Technical Summary

Problems solved by technology

Users use the friend search function to find friends to expand their social circle, but this search wastes a lot of time, is inefficient and not accurate

The major social networking platforms have successively launched various recommendation strategies to meet user needs, but the current recommendation algorithms have many limitations.

The classic collaborative filtering method cannot solve the problem of data sparsity very well, and considers less user information, so the recommendation effect is not ideal

The tag-based recommendation method focuses on the user's static attributes and ignores the more socially valuable dynamic interaction behavior, so that the characteristics of social network users cannot be well obtained, and the recommendation effect is not ideal.

[0004] The existing friend recommendation method considers the problem of simplification, does not conduct a comprehensive analysis of user characteristics, and has limitations to varying degrees. Therefore, the present invention proposes a method that comprehensively considers the various characteristics of social network users and infers user portraits. So as to accurately personalize the friend recommendation method

Images