基于改进LDA模型的图书推荐方法研究

Book Recommendation Method Based on Improved LDA Model

当前海量的商品信息困扰着顾客,商用图书推荐系统亟待研究与应用。本文提出一种改进LDA模型的图书推荐方法,根据用户微博内容和图书描述分别生成主题模型,并基于主题的词分布将表示用户主题的向量转化成由图书主题构成的向量,从而更精确地计算用户与图书之间的相似度,最后排序得到推荐结果。在实验中,本文对新浪微博用户推荐亚马逊图书,并与传统方法进行了对比。实验结果表明本文的方法效果更佳,可为当前的图书推荐提供参考。

Massive commodity information is puzzling customers, and commercial book recommendation system urgently needs to be studied and applied. This paper proposes an improved LDA model recommendation method. It generates topic model based on user miroblog contents and book descriptions, translates words that represent user topics into words whick represent book topics by means of the word distribution, and calculates the similarity between users and books to get the ranking of recommended results. In the experiment, this paper recommends Amazon books to Sina Microblog users, and compares it with the traditional methods. The experimental results show this method presented better performances, which could provide reference for current book recommendation.