Research On Personalized Recommendation Method Based On Topic Model

The personalized recommendation system analyzes the user's historical behavior data,establishes a corresponding interest preference model,predicts its personalized needs,generates recommendations,and is widely used in e-commerce websites and social platforms.User needs are divided into explicit and implicit.Constructing effective and appropriate models for different user needs will help improve the accuracy of personalized recommendations.When facing the explicit needs of consumers,this paper proposes a hierarchical Dirichlet process mixture model and personalized Pagerank algorithm to model the unstructured information of the product.First,The HDP model is proposed to model the descriptive text information of the product,and the potential product function topics are inferred for clustering.Then use the personalized PageRank algorithm to model the tag information and generate a recommendation list for each target product.Experiments on real cloud service product data sets show that the proposed HDP model performs well in clustering effects,and the two-stage recommendation model based on the HDP model and personalized PageRank has a greater improvement in recommendation performance than other methods,and effectively alleviates the cold start problem of new products.When facing the implicit needs of consumers,this paper uses the topic model to model the user's historical purchase behavior,introduces the concept of latent group,and divides the users into non-overlapping groups with similar preferences.At the same time,this paper defines the double sparse concept of the latent group interests and interest products,then construct a dual-sparse generative model based on latent groups,and use group interest to predict individual interest.The experimental results on the Taobao user purchase behavior data set show that the proposed model can effectively alleviate the sparseness problem of individual users and improve the accuracy of individual demand prediction.At the same time,the proposed double sparse strategy is more in line with reality.