基于主题模型包含突发因素的推荐算法研究

Research on Recommendation Algorithm Based on Topic Model Containing Bursty Factors

推荐系统是社交平台个性化服务的重要工具,协同过滤算法由于其推荐的准确性和高效性已经成为推荐领域最经典的算法之一。论文提出一种结合突发话题检测和主题模型的混合协同过滤方法。该算法在语料筛选阶段加入了突发因素,使通过主题模型LDA话题训练的话题具有时效性,然后在低维主题-文档概率分布上计算用户和项目的相似度;最后采用邻域方法预测未知评分。实验表明,该方法适用于微博突发话题的推荐,显著提高了推荐系统的时效性和准确性。

The recommendation system is an important tool for the personalized service of the social platform. The collaborative filtering algorithm has become one of the most classic algorithms in the recommendation field because of its accuracy and efficiency. This paper proposes a hybrid collaborative filtering method combining burst topic detection and topic model. The algorithm adds sudden factors in the corpus screening stage,which makes the topic trained by the topic model LDA topic time-sensitive,and then calculates the similarity between users and projects on the low-dimensional topic-document probability distribution. Finally,the neighborhood-based approach is used to predict the unknown score. Experiments show that the method is suitable for the recommendation of microblogging bursty topics,which significantly improves the timeliness and accuracy of the recommendation system.