融合LDA主题和Doc2vec算法的DeepFM模型的推荐算法研究

Research on Recommendation Algorithm of DeepFM Model Integrating LDA Topic and Doc2vec Algorithm

如今,有很多辅助决策算法在日常生活的各个方面为人们推荐个性化内容或产品。本文以医疗信息推荐作为案例,研究提出一种融合狄利克雷分配(Latent Dirichlet Allocation,LDA)主题模型和Doc2vec算法的DeepFM模型。该模型能够挖掘评论文本中的隐藏主题和隐藏特征并考虑隐藏特征的交叉情况,能够在保留评论文本表层信息的同时学习数据中的浅层和深层特征。本文将该模型与之前的模型在真实的数据上进行实验对比。实验结果表明,相较于现存模型,该模型的推荐准确率有了一定的提高。

Nowadays,there are many assistant decision-making algorithms to recommend personalized content or products for people in all aspects of daily life.Taking medical information recommendation as a case study,this paper proposes a DeepFM model which integrates the topic model of Latent Dirichlet Allocation(LDA)and Doc2vec algorithm.The model can mine the hidden topics and hidden features in the comment text,consider the intersection of hidden features,and learn the shallow and deep features in the data while retaining the surface information of the comment text.This paper compares the model with the previous model on real data.The experimental results show that compared with the existing models,the recommendation accuracy of this model has been improved to a certain extent.