提高会话推荐多样性的SRL推荐系统模型

现有的各种推荐多样性定义难以体现用户在会话期内推荐项目的多样性，且现有的各种提高推荐多样性的方法通常以牺牲推荐准确率为代价.提出了会话推荐多样性的概念，分析了现有推荐系统会话推荐多样性偏低的主要原因是推荐树中存在过多的重复节点.设计了一种会话推荐列表SRL(sessionrecommendationlist)推荐系统模型，为每个活跃用户建立会话推荐列表，能够有效避免推荐树中出现推荐环路或弱推荐环路.通过MovieLens数据集测试表明，SRL推荐系统模型可以大幅度提高会话推荐多样性，同时也提高了推荐准确率.

SRL Recommendation System Model Improving Session Recommendation Diversity

Current diversity definitions cannot indicate the recommendation diversity during users' session and the existing methods for improving recommendation diversity always come at the expense of precision. Session recommendation diversity was proposed. Traditional recommendation systems have poor session recommendation diversity because there are too many repeated nodes in the recommendation trees. SRL model was designed to eliminate the redundancy. By creating session recommendation list for each active user, recommendation loops or weak recommendation loops in the recommendation trees could be avoided based on the proposed model. Experimental results on MovieLens dataset showed that SRL model has substantially higher session recommendation diversity and better recommendation precision.