基于项目权重的协同过滤推荐算法研究

Research on Collaborative Filtering Recommendation Algorithm Based on Item Weight

推荐系统的主要功能是挖掘用户的行为信息,主动向用户推荐其感兴趣的内容,其中推荐算法作为推荐系统实现的核心而得到广泛关注。目前,主流的推荐方法有两种:基于协同过滤的推荐和基于内容的推荐。本文首先对基于用户的协同过滤方法进行较为深入和全面的分析,进而通过增加冷热门权重项对传统的余弦相似度计算方法进行改造。根据在MovieLens数据集上的实验研究表明,将上述改进应用于基于用户协同的过滤算法,能够提升推荐系统的精确率。

The main function of recommender system is to mine users*behavior information and actively recommend their interested content to users.As the core of recommendation engine,recommendation algorithm has been widely studied.At present,there are two mainstream recommendation methods:collaborative filtering based recommendation and content-based recommendation.Through a more in-depth and comprehensive analysis of the user based collaborative filtering method,the traditional cosine similarity calculation method is improved by adding hot and cold weight items.According to the experimental research on movielens dataset,the application of the above improvements to the user based collaborative filtering algorithm can improve the recommendation effect to a certain extent.