Temporal-based ranking in heterogeneous networks

Abstract

Ranking is a fundamental task for network analysis, benefiting to filter and find valuable information. Time information impacts results in content that is sensitive to trends and events ranking. The current ranking either assumes that user's interest and concerns remain static and never change over time or focuses on detecting recency information. Meanwhile most prevalent networks like social network are heterogeneous, that composed of multiple types of node and complex reliance structures. In this paper, we propose a general Temporal based Heterogeneous Ranking (TemporalHeteRank) method. We demonstrate that TemporalHeteRank is suitable for heterogeneous networks on the intuition that there is a mutually information balance relationship between different types of nodes that could be reflected on ranking results. We also explore the impact of node temporal feature in ranking, then we use the node life span by carefully investigating the issues of feasibility and generality. The experimental results on sina weibo ranking prove the effectiveness of our proposed approach. © 2014 IFIP International Federation for Information Processing.