基于项目语义相似度的协同过滤推荐算法

Collaborative Filtering Recommendation Algorithm Based on Semantic Similarity Between Items

协同过滤是个性化推荐系统中最广泛使用的推荐技术。在用户评分矩阵极度稀疏情况下,传统的协同过滤推荐算法中用户相似度的计算建立在用户评分项目交集之上,并且没有考虑不同项目之间存在的语义关系,从而导致推荐准确率低。针对上述问题,文章提出一种新的基于项目语义相似度的协同过滤算法(CFSSI,collaborative filtering based on semantic similarity between Items):首先利用领域本体计算项目之间的相似性,填充评分矩阵缺失值,而后根据修正的余弦相似度计算用户相似性。实验结果表明:算法可以在用户评分数据极端稀疏的情况下,仍能取得较高的推荐质量。

Collaborative filtering is one of the most popular technologies in the personal recommendation system. With user rating data becoming extremely sparse, traditional collaborative filtering recommendation algorithm calculates similarity between users using the intersection of different user rating items, and it does not consider the semantic relationship between different Items, thus recommendation quality is very poor. In order to avoid the problems above, a novel collaborative filtering algorithm based on semantic similarity between items is presented. Firstly, this method calculates similarity between items according to domain ontology, fills user rating matrix, and calculates users＇ similarity with adjusted cosine measure. The experiment result shows that this method can effectively improve recommendation quality even with extreme sparse of user rating data.