Personalized Information Recommendation Based On Random Walk Model

With the rapidly growing amount of information available on the WWW, it is difficult to find desired information from huge amounts of content. Therefore, it becomes necessary to have tools to help users to select the relevant part of online information. To satisfy this need, recommender systems have emerged and they have been applied successfully in many websites, e.g. YouTube, MovieLens, Amazon. Collaborative filtering is the most popular approach to build recommender systems and has been successfully employe in many applications. However, data sparseness and cold start problem are the two major Barriers affecting the accuracy of the recommendation results. Collaborative Filtering methods make recommendations only based on the users'ratings. With the advent of online social networks, the trust-based approach to recommendation has emerged.To alleviate the sparsity problem in collaborative filtering, this paper proposes an algorithm named item-tag-based random walk recommender (TRWR), which is based on the random walk recommender (RWR) algorithm. The idea in this paper is as follows:firstly, for a target user, it calculates transition probabilities between items and tags separately, then takes finite length random walks in the item space and the tag space in order to generate some recommended items. Secondly, it recalculates these items'scores. Finally, it recommends some items to the target user.When calculating similarities between objects, we introduce the number of same scorers. Experiments show that this method is better than Cosine and Pearson Correlation.Besides ratings information and tags information, we introduce a trust network among users into personalized information recommendation. This trust-based approach can make better recommendation for so-called cold start users that have rated only a very small number of items.In conclusion, we use the random walk model to make recommendation, utilize more information about users, and alleviate the sparsity problem and the cold start problem. We performed evaluations on the MovieLens dataset and Epinions dataset, and compared our approaches with existing recommendation methods.