Link Prediction Using Double Degree Equation with Mutual and Popular Nodes

Abstract

Link prediction among the objects of data is one of the essential tasks in several application fields, such as spam mail detection, expert detection, information retrieval, influence detection, recommender system, surveillance of communication, and disease prediction. Link prediction aims to estimate the missing link based on the existing links in the given complex network. To solve the link prediction problem, different techniques have been introduced. The similarity-based link prediction techniques are effective, but still imperfect. To better the predicting accuracy, we introduce a link prediction technique that aims to improve the accuracy of the existing link prediction problem by combining the mutual nodes or neighbor with the feature or information of popular nodes. We also examine the proposed technique empirically and lead extensive tests over complex network datasets obtained from different domains. The experimental results show that the proposed technique has performed well as compared to other state-of-the-art techniques. The average result of the proposed link predictor is 0.8812 while others are CN 0.8734, Katz 0.8112, LHN 0.7778, PA 0.7367, and PD accuracy is 0.6538.