Link Prediction Method for Social Network

Abstract

Aiming at the problem of low prediction performance due to the lack of negative link feature fusion and mining effective information in symbolic social networks, a new negative link prediction method based on feature fusion is proposed in this paper. Based on the classical structural balance theory and social status theory, this method constructs four characteristics related to negative symbols. Including node characteristics, structure characteristics, similarity characteristics and scoring characteristics, the negative link prediction is realized by using logistic regression algorithm. Its effectiveness is verified on two typical symbolic network data sets, epinions and Slashdot. The experimental results show that compared with the benchmark method, the accuracy of the extracted method is improved by about 4.5% and 10.4% respectively on the two data sets. The F score increased by about 27.3% and 31.5% respectively. The purpose of improving the prediction effect of negative link is achieved..