语义相似性与协同过滤集成推荐算法研究

基于项目协同过滤算法能提高基于用户协同过滤方法的扩展性问题,并考虑项目之间的关系避免计算用户之间关系的瓶颈,但基于项目协同过滤算法依然存在稀疏性和新项目预测等问题。为了解决这些问题,该文采用了一种基于项目的结构化语义信息的集成相似性算法。为了抽取项目的语义信息,通过本体学习建立特定领域本体并利用包装器代理从网站中抽取本体类的实例和项目属性。实验结果证明了此方法不仅能很好的解决基于项目协同过滤算法带来的问题,而且还提高了推荐精度。

Research of Recommendation Algorithm on Integration of Semantic Similarity and the Item-based CF

Item-based Collaborative Filtering algorithms can enhance the scalability problems associated with traditional user-based Collaborative Filtering approaches and avoid the bottleneck of computing user-user correlations by considering the relationships among items.But it still worked poor in solving the problem of sparsity,predictions for new Items.In order to resolve efficiently several problems,this paper introduced an integrated similarity algorithms based on structured semantic knowledge about Items.We built domain-special ontology by ontology learning and used wrapper agents to automatically extracting instances of the ontology classes and semantic properties about Items from web site.Experimental results showed that the integrated similarity algorithms efficiently deal with the problems associated with Item-based Collaborative Filtering algorithms as well as improving accuracy.