A Collaborative Filtering Recommendation Algorithm Based on User Preferences on Service Properties

In the service recommendation, the data of user ratings are usually very sparse. In the case of data sparsity, item similarity which is based on user ratings in the traditional item-based collaborative filtering algorithm ignores the situation that users are different in regarding various item properties. That results in the low accuracy when predicting. Based on this point, this paper proposed a collaborative filtering algorithm based on user preferences on service properties to solve the data sparsity problem in the service recommendation. This method firstly builds the service property preference model for each user based on information theory. Secondly, computes the service similarity correction factors of each user on any two services with service properties similarity. And finally the similarity of two services is the sum of service similarity correction factor and the Pearson correlation coefficient of them. The experiment results suggest that the proposed algorithm can efficiently improve the recommendation accuracy in the case of data sparsity.