A Network Information Filtering Method Based on Node Energy Transfer

Abstract

Network information filtering can be projected into link prediction in bipartite network to alleviate information overload. The paper discovers that node transfers redundant recommendation energy in network and gives bad effect on information filtering efficiency. Gathering node’s overlapping degree by factorizing negative matrix, redundant energy is reduced with overlapping degree and initial recommendation energy. The method endows information filtering system with dynamic recommendation prediction, the empirical results indicates the method can enhance information filtering performance.