Item Type Based Collaborative Algorithm

Due to the high sparseness of data, traditional collaborative filtering algorithms suffer from bad scalability and inaccuracy. In this paper, we proposed a new algorithm to avoid the negative effect of the high sparse data and to improve the accuracy of traditional collaborative filtering recommendation algorithms. In our algorithm, the types of rated terms by a user are checked first in order to decide the favorite types to be recommended, and then the nearest neighbors of non-rated items in these favorite types will be computed. According to the neighbors, ratings will be evaluated and added for the non-rated items. And then, the nearest neighbors of a user can be calculated based on the new ratings. Finally the recommendation to the user can be made based on the neighbors. Empirical results show that our proposed algorithm has a lower mean absolute error (MAE) than other traditional algorithms. Our algorithm can effectively overcome the sparseness of data and perform better than traditional collaborative filtering algorithms.