基于物品属性聚类的融合协同过滤算法

With the development of information technology, a lot of recommendation systems have been widely used in e-commerce, news recommendation etc. Collaborative filtering is one of the most important recommendation algorithms among recommendation system. However it's also suffering some problems such as data sparsity and cold start. In order to reduce computing time for recommendation and improve the accuracy, a new collaborative filtering recommendation system combining item-tag clustering and slope-one algorithm is proposed. In this algorithm, users were clustered according to users' preference on tags of items. Property tags were userd to reduce users' feature vector, which solved data sparsity partially. By combining both clustering and slope-one algorithm, the recommendation system retained the advantage of slope-one algorithm while improving accuracy as well. The experiment were applied to MovieLens dataset, which showed that the accuracy of this combined algorithm is in advance of na?ve slope-one.