A Link Prediction Algorithm for Weighted Networks Based on Dempster-Shafer Evidence Theory and Node Multi-features

Existing link prediction algorithms for weighted networks always only consider the impact of a single factor on the similarity. The similarity index based on local attributes has a coarser granularity. So when the network structure is incomplete, the accuracy is low. Meanwhile the computational complexity of extracting higher-order paths is relatively high. To solve the above problems, a link prediction algorithm for weighted networks combining Dempster-Shafer evidence theory and node multi-features is proposed. It makes full use of the node's degree, strength, edge weight, path information, triangular feature, and other information. And by introducing the concepts of weighted common neighbor node clustering coefficient and weighted path clustering coefficient, a new measure of similarity based on multi-source information fusion rule in Dempster-Shafer evidence theory is finally obtained. Experiments are carried out on eight real weighted datasets. The AUC and Precision are used as evaluation indicators to compare and analyze with the classic similarity indicators of WCN, WAA, WRA, and weighted Jaccard, as well as five recent algorithms. The results show that the prediction performance of the proposed algorithm is better than other algorithms.