Applying Bi-directional Link Mining in Personalized Recommendation

Abstract

Recently, many researchers have been attracted in link prediction which is an effective technique to be used in graph based models analysis. By using link prediction method we can understand associations between nodes. To the best of our knowledge, most of previous works in this area have not explored the prediction of links in dynamic Multi-dimension Networks and have not explored the prediction of links which could disappear in the future. We argue that these kinds of links are important. At least they can do complement for current link prediction processes in order to plan better for the future. In this paper, we propose a link prediction model, which is capable of predicting bi-direction links that might exist and may disappear in the future in dynamic Multi-dimension Networks. Firstly, we present the definition of multi-dimensional networks, reduction dimension networks and dynamic networks. Then we put forward some algorithms which build Multi-dimension Networks, reduction dimension networks and dynamic networks. After that, we give bi-direction link prediction algorithms in dynamic multi-dimension weighted networks. At the end, algorithms above are applied in recommendation networks. Experimental results show that these algorithms can improve the link prediction performance in dynamic multi-dimensional weighted networks.