Symbolic link prediction method based on hidden space

Abstract

The invention relates to a symbolic link prediction method based on latent space. The present invention calculates the weight between nodes in the signed network according to the common neighbor index and the adjacency matrix of the signed network, and obtains the weight matrix w. To obtain the hidden space matrix H and the corresponding mapping matrix u that need to be optimized at the end of the method, they are respectively fixed Hp, up, Hn, un among the three items, find out the optimal iterative formula of Hp, up, Hn, un, and iteratively update Hp, up, Hn, un at the same time according to the optimized iterative formula obtained, according to the Hp after iteration , the optimal value of up, Hn, un, to make predictions on the possibility and sign of link existence. The invention overcomes the defect that previous studies only focused on linking symbols, but did not predict the possibility of the existence of the symbols. The present invention proposes a hidden space-based symbolic link prediction method for the symbolic prediction problem of a symbolic social network, which has a fast convergence speed and can make more accurate predictions about the symbol of the link and the probability of being the symbol.

Description

technical field [0001] The invention belongs to the analysis of link prediction methods applied in signed social networks, in particular to a hidden space-based symbolic link prediction method. Background technique [0002] With the rapid rise of a series of social networking sites, such as Facebook, Twitter, LinkedIn, Epinions and so on. In order to ensure the user experience, a lot of energy has been devoted to the research of social mechanism. Traditional social network analysis mainly only considers unmarked social networks such as Facebook and MySpace. These networks can be transformed into graph models, in which nodes represent entities, and positive weighted edges represent whether there is a relationship between entities and this relationship importance. Recently, research on signed directed social networks is emerging, in which relations between users can be either positive (indicating trust between users) or negative (indicating user The relationship between is ...

AI Technical Summary

Problems solved by technology

At present, these methods only focus on the research of link symbols, but do not predict the possibility of the existence of the symbol. How to design a method that can not only predict the symbol but also obtain the probability of the existence of the symbol is the main problem at present.

Images