A novel hybrid algorithm of personalized link prediction

Link prediction is an important problem in the traditional field of data mining. Thus in the past several years, abundant traditional algorithms from a global perspective have been proposed. But, they all fail to accurately and individually recommending friends in social networks. Hence, the personalized link prediction has been introduced in this paper, in which each individual will get equal number of diversiform predictions. Inspired by previous researches in other fields, heat conduction process will be generalized to improve the performance of prediction. Then one better hybrid algorithm composed of local random walk and heat conduction has been proposed. This algorithm can outperform other algorithms simultaneously in all widely-used metrics: auc, precision and recall. In general, this work may shed some light on the in-depth understanding of the application of network sciences in data mining.