Link prediction methods and their accuracy for different social networks and network metrics.

Abstract

Currently, we are experiencing a rapid growth of the number of social–based online systems. The availability of the vast amounts of data gathered in those systems brings new challenges that we face when trying to analyse it. One of the intensively researched topics is the prediction of social connections between users. Although a lot of effort has been made to develop new prediction approaches that could provide a better prediction accuracy in social networked structures extracted from large–scale data about people and their activities and interactions, the existing methods are not comprehensively analysed. Presented in this paper, research focuses on the link prediction problem in which in a systematic way, we investigate the correlation between network metrics and accuracy of different prediction methods. For this study we selected six time–stamped real world social networks and ten most widely used link prediction methods. The results of our experiments show that the performance of some methods have a strong correlation with certain network metrics. We managed to distinguish ’prediction friendly’ networks, for which most of the prediction methods give good performance, as well as ’prediction unfriendly’ networks, for which most of the methods result in high prediction error. The results of the study are a valuable input for development of a new prediction approach which may be for example based on combination of several existing methods. Correlation analysis between network metrics and prediction accuracy of different methods may form the basis of a metalearning system where based on network characteristics and prior knowledge will be able to recommend the right prediction method for a given network at hand.