协同过滤算法的改进研究

The Improved Research of Collaborative Filtering Algorithm

随着互联网的高速发展，网络信息的规模越来越大，为了更高效的为用户提供信息服务，运用协同过滤算法的个性化推荐技术正广泛的应用于各类网络信息服务系统。然而协同过滤算法存在着冷启动、数据稀疏等问题，影响了系统的推荐精度。为了改善这一问题，本文在对传统的协同过滤算法分析研究的基础上，提出了一种改进方案，将基于项目的协同过滤与基于用户的协同过滤技术联合运用，通过相关反馈和查询扩展的信息扩展方法，在一定程度上缓解了用户一项目评分矩阵数据稀疏性问题，与传统算法相比提高了推荐准确度。

With the rapid development of the Internet, there are more and more network information. In order to provide a more ef- ficient information services, collaborative filtering based on personalized recommendation technology is widely used in various types of network service system. However, there are some problems with collaborative filtering algorithms, such as cold start, data sparse, these affect the accuracy of the information which systems is recommended. In this paper we studied the principles of col- laborative filtering algorithm, proposed an method to improve the accuracy. It uses item-based collaborative filtering with user- based collaborative filtering, expands the information by relevance feedback and query methods, and it raise the recommendation accuracy than traditional algorithms.