A novel parameter free approach for link prediction

Social networks have become the most common way to connect people and to diffuse information, thus many efforts have been devoted to analyze their mining and evolution. The link prediction is one of the most important issues in the computational analysis of social networks. The main goal is to predict links that may appear in the future. To this purpose, a new similarity based method is proposed whose main strategies rely on the path depth from a source node to a destination node and their degrees. The experimental results on five instances of social networks show the impact of the path length on the method performance. Based on the values of the area under curve (AUC), the proposed method predicts links with high accuracy when compared with the existing ones.