摘要
情境对于推荐系统的重要性已经得到了众多学者的普遍认可。然而在现有的基于情境感知的推荐中,基本上赋予所有情境因素的权重都是一样的,这种权重等分的做法很大程度上影响了推荐结果的质量。因此,提出了一种用户情境下基于信息增益和项目的协同过滤推荐技术,运用信息增益理论对诸多情境因素进行属性约简,计算不同情境属性的权重,抽取出对推荐结果影响较大的重要情境信息,将其与传统的基于项目的协同过滤推荐算法相结合,为处于特定情境下的用户提供个性化推荐。最后,通过实验证明该技术可以有效地提高推荐结果的准确率。
The importance of context in recommender systems has gained general recognition of the numerous scholars. However, in the existing context-aware recommendation, the weights of all the contextual factors are basically the same, which largely limits the quality of the recommended results. Therefore, this paper proposes a context-aware recommendation technology based on information gain and item-based collaborative filtering, applies the theory of information gain to reduce contextual factors, calculates the weights of different contex-tual attributes and extracts the significant contextual information that influences the recommended results greatly. The combination with the traditional item-based collaborative filtering algorithm provides appropriate items to specific users under particular contexts. Finally, the experiment shows that the proposed approach is helpful to improve the accuracy of recommended results.
出处
《情报杂志》
CSSCI
北大核心
2014年第7期165-170,共6页
Journal of Intelligence
关键词
个性化推荐
情境感知
信息增益
属性约简
协同过滤
personalized recommendation
context-aware
information gain
attributes reduction
collaborative filtering
