关键词: 协同过滤, 平均绝对误差, 均方根误差, 用户关系网, 社区划分

Abstract:

In order to raise the accuracy and extensibility of collaborative filtering recommendation,a collaborative filtering recommendation based on network community is proposed. First, a network of users could be built by the metadata about friends. Then the network of users could be divided by community partition algorithm to make sure that the same community of users have similar interests. The nearest neighbour set of target users are found by users of the same community. Finally, scores of target users are forecasted according to neighbour users'scoring to unknown projects. Results show that when the number of neighbours is less than 27, the accuracy got by the proposed algorithm is better than the algorithm based on user fuzzy clustering.

Key words: collaborative filtering, MAE, RMSE, user network, community division