一种混合模式的协同过滤算法

当今是一个数据爆炸时期，促进信息过滤技术发展，个性化推荐系统作为其中一种重要的应用方式，已经成为很多网站一种个性化信息服务方式，但传统的协同过滤算法存在扩展性和稀疏性的问题。提出一种基于项目聚类、项目语义相似度和奇异值分解的混合推荐模型，来应对传统的协同过滤推荐系统面临的算法的伸缩性问题、数据稀疏性问题和推荐的精准度问题，进行推荐。结果表明，与传统的算法相比，使用该改进算法能显著地提高推荐系统的推荐质量。

A Mixed Mode Collaborative Filtering Algorithm

The world has entered a data explosion era, promoting the rapid development of information filtering technology. Personalized recommen- dation system is an important form of information filtering, has become indispensable to each big mainstream website of a new generation of personalized information service form. To solve the traditional collaborative filtering algorithm existing the sparsity and sealability prob- lems, proposes a collaborative filtering algorithm based on Canopy clustering, Semantic similarity between items and singular value de- composition to deal with the traditional collaborative filtering recommendation system scalability, the algorithm of data sparsity problem and recommendation accuracy problem. The experimental results show that, compared with the user based collaborative filtering algo- rithm, this algorithm can improve the similarity calculation accuracy under the cold start problem.