基于用户行为模型和蚁群聚类的协同过滤推荐算法

协同过滤技术是推荐系统中应用最为广泛的技术之一,用户的相似性度量是整个算法的核心要素,会对推荐算法准确率产生很大的影响.传统的协同过滤算法过度依赖用户评分机制,影片自身的标签信息没有被考虑为一个影响因素,在用户聚类时采用K近邻算法,会由于评分矩阵过于稀疏而难以收敛.同时,传统推荐技术仅基于用户历史行为进行推荐,无法为新用户提供合理的推荐.针对以上问题,提出了一种基于用户行为建模的蚁群聚类和协同过滤算法相结合的影片推荐技术.

Collaborative Filtering Recommendation Based on Ant Colony Clustering and User Behavior

Collaborative Filtering Recommendation （CFR） is one of the most widely used technologies in recommendation systems, in which the measurement of user similarity is the core element and can affect the accuracy of results a lot. Traditional CFR only depends on the movie ratings of users. As a result, it generates very different results when different individuals rate movies according to their personal standards. The performance of K-means clustering method is hard to converge when the matrix of user ratings is very sparse. As the traditional CFR only depends on the rating history of users, when a new user registers in the system, traditional CFR can hardly give desirable results. It cannot recommend movies to users according to the movie which the target user is watching or just watched, either. To solve the above problems, a kind of algorithm is proposed based on Ant Colony Clustering and user beha- vior to combine with traditional CFR algorithm.