Performance evaluation of baseline link prediction techniques on simple and complex networks

Abstract

The world wide web, the internet, social interacting entities, and neural networks are the highly interconnected systems and are considered as complex networks. Now a days, scientists are focusing on link prediction in social networks. In literature survey, many supervised and unsupervised algorithms have been proposed for link prediction in simple and complex social networks. This article summarized the recent work about link prediction in complex social networks. Three data sets have been taken and baseline predictor methods common neighbor, jaccard coefficient, adamic/adar, preferential attachment, LRW and SRW are implemented on these datasets. Analysis of the results is done with the experiment by calculating two standard metrics AUC and precision.