基于MapReduce的混合推荐算法及应用

Hybrid Recommendation Algorithm Based on MapReduce and Its Application

针对基于项目与基于用户两种传统协同过滤算法的不足,文中结合基于用户以及基于项目的两种传统协同过滤算法,并加以合理改进,提出了一种新型的混合型并行推荐算法。通过对新算法MapReduce编译,使新算法能够在Hadoop云平台下顺利运行。在可以利用以基于用户的方法为基础划定出定量的邻居范围,保证了推荐的个性化,同时,利用基于项目的协同过滤算法进行推荐,最终根据综合因素调整评分预测方法得出符合实际的推荐结果。实验结果表明,在数据量相对较大时新算法不仅在处理速度上表现更加优越,而且明显提高了推荐精确度。同时文中将该算法应用在西安本土旅游推荐服务上,针对西安市几大景点进行推荐,使新算法的准确性在实际应用中得到验证。

For the shortcomings of traditional project- based and user- based collaborative filtering algorithm,a newparallel recommendation algorithm is proposed,combined user- based with project- based collaborative filtering algorithm and improved them. Through MapReduce compilation,the newalgorithm can run in Hadoop cloud platform. To guarantee the personalized recommendation,it can take advantages of the collaborative filtering algorithm based on user defined a certain number of neighbors. At the same time,the project-based collaborative filtering algorithm is used to recommend. Finally,according to the comprehensive adjusted score prediction method,the recommended results are obtained. The experimental results showthat the algorithm becomes more superior in the case of a large number of processing speed,and improves the accuracy of recommendation. Simultaneously,the algorithm is applied in local tourism of Xi＇an referral service for several major attractions to recommend. The accuracy of the newalgorithm has been verified in practical applications.