Link prediction method in large-scale microblog heterogeneous information network

Abstract

The invention relates to the field of Internet technology, and provides a link prediction method in a large-scale microblog heterogeneous information network. The link prediction method comprises the following steps: filtering users according to preset strategy; extracting a plurality of links from the network, wherein a positive example set is ET, and a negative example set is EF; calculating the characteristics of all nodes in ETUEF and the characteristics of the links in an E-ET-EF network, and converting the characteristics of the nodes into the characteristics of link relations; dividing the EFUET into a training set, a verification set and a test set, training models on the training set, selecting model hyper-parameters with the optimal prediction result on the verification set to obtain a final model h theta (x) and a threshold value theta; putting any link relation in the test set into the model, so that the probability P generated by the link relation can be obtained. Experiments show that the area under curve and F value of the method provided by the invention are obviously improved compared with a method based on local information similarity and path similarity, and the method has higher maximum K accuracy rate stability.

Description

technical field [0001] The invention belongs to the technical field of the Internet, and in particular relates to a link prediction method in a large-scale microblog heterogeneous information network. Background technique [0002] At present, with the rapid development of mobile Internet and the wide application of search engines, portal media, social networks, etc., the Internet has gradually become a platform containing massive information. Sina Weibo is the most widely used microblog system in China. Since its launch in 2009, the number of registered users has exceeded 500 million. In the microblog system, users can send microblogs (similar to a message, with a word count of less than 140), comment on microblogs, forward microblogs, etc. The link relationship in Weibo includes friend relationship, follow relationship, @ relationship and so on. These relationships are all directed and can be represented as a directed graph. Microblog is a typical heterogeneous informati...

AI Technical Summary

Problems solved by technology

[0006] (2) Important laws and phenomena in sociology, such as the Matthew effect, the 28th law, etc., are difficult to be simply represented by similarity

However, these similarities can only characterize certain aspects of the network, and have no proven direct correlation with the generation of new link relationships between network nodes

Images