基于加权Slope One的协同过滤个性化推荐算法

摘要

The traditional CF algorithm has the problem of cold start,data sparseness and low operation efficiency.This paper chose Slope One algorithm which was more efficient and accurate than traditional CF algorithm for research and analyses its advantages, principle and process.It pointed out shortcomings of Slope one algorithm that it did not take user interest changes and user similarity into account.This paper put forward improved scheme of Slope one algorithm based on user interest forgetting function and user nearest neighbors.The experimental test on MovieLens dataset proves the feasibility and better time performance of improved scheme.%针对传统协同过滤算法存在冷启动、数据稀疏、运行效率低下等问题,分析了较传统协同过滤算法更加高效准确的Slope One算法的优点、原理及流程.针对Slope One算法未考虑用户兴趣变化和用户相似性这两方面的问题,提出了基于用户兴趣遗忘函数和用户最近邻居筛选策略的改进方案,以期提高推荐的质量,同时采用MovieLens数据集进行了实验验证.实验对比结果佐证了该算法确实提高了推荐准确度并且减少了响应时间.