Structure Similarity And Its Applied Research, The Recommended System

With the rapid development of the network and E-commerce, we have entered the era of information explosion. It makes people's life more and more different from each other. But at the same time, it also brings the problem of information overload. Facing the massive information, on the one hand users can not find the information they want effectively, on the other hand much information which no people have cared are lost. In this case, personalized recommendation system appears.The main aim of personalized recommendation system is to provide users the information which can meet their different interests. That is to say, the nature of personalized recommendation system is information filtering. Especially, it is also very important in e-commerce field. Collaborative filtering is one of the most successful recommendation techniques, which can effectively predict the possible future likes of users based on their past preferences. The key problem of this method is how to define the similarity between users. The goal of this paper is to analyze and compare the performances of different kinds of similarity indices in recommendation system.The main work of this paper is just as below:Firstly, we have researched deeply in personalized recommendation system and analyzed the content, the range and the status of different kinds of recommendation technology. We discussed collaborative filtering which is one of the most popular personalized recommendation systems. Then the paper has introduced data mining, complex network and link prediction.Secondly, we introduced six structure-based similarity indices and compare their performances in collaborative filtering system with two benchmark indices, Cosine index and Pearson similarity. The structure-based similarity indices are based on the relation of the points in the network to define the similarity between users or items. What is more, we have done several groups of experiments to compare their performances in collaborative filtering system.Thirdly, on the base of the two different kinds of similarity approachs we have propesed an improved structure-based similarity method which includes network structure and ratings information. So it can avoid the limit which was brought by the two different kinds of similarity approachs.At last, with several performance indexes, we have proved that the structure-based similarity in the paper can effectively enhance the quality and the effect of the recommendation system.