Abstract: In view of the fact that background data source of recommended objects is single and commodity recommendation lacks diversity in traditional recommendation systems, this paper puts forward a diversified commodity recommendation algorithm. The relationship between objects is divided into three categories, which is respectively complementary correlation, similarity correlation and scenario correlation. It analyzes these correlations between customer interest ontology and recommended commodities, as a result, deduces commodity recommendation list of the customer. The experimental results show that compared with traditional recommendation method only to consider similarity correlation, the algorithm enriches the type of recommended commodities, optimizes recommendation list and further meets consumers’ demand for complementary or scene correlation commodities.

Key words: semantic association, commodity recommendation, ontology, relational data

摘要： 针对社会化商品推荐中推荐对象背景数据单一、推荐结果缺乏多样性等问题，提出了一种基于推荐对象间关联关系的多样性推荐算法。在领域本体建模的基础上，将推荐对象之间的关联分为三类，即互补关联、相似关联和情景关联，分析推荐商品与消费者兴趣本体之间的综合商品相似度，综合商品互补度和商品情景关联度，最后根据算法得出各商品推荐度及推荐列表。实验结果表明，该方法与传统的推荐方法相比，在一定程度上丰富了推荐商品的类型，优化了推荐列表排名，进一步满足了消费者对互补性商品及情景关联性商品的推荐需求。

关键词: 语义关联, 商品推荐, 本体, 关联数据