Research On Hybrid Recommendation Algorithm Based On Deep Learning

With the rapid development of Internet technology,the amount of data generated every day is exploding.Today,with the heavy overload of information and the extremely competitive competition,traditional search engines are no longer able to meet current needs.The recommendation system emerged in this situation and became a very important part of many websites.It has also become a new darling of the new era.The most widely used one is the collaborative filtering technology.Although the collaborative filtering algorithm should be the most extensive,it also faces serious problems such as data sparsity,scalability and cold start.In order to solve these problems,this paper proposes a regression-based conditionally constrained Boltzmann machine recommendation model(R-CRBM),as well as a reinforcement-based restricted Boltzmann machine model(S-CRBM)and a hidden factor model(LFM)Hybrid Recommendation Model(SCRBM-LFM).Firstly,this paper introduces a conditional layer based on the RBM model,and uses the user information and project information to train the model,and uses the linear return algorithm to fuse the two results.Secondly,based on the condition-based restricted Boltzmann machine model,the reinforcement layer is introduced.The reinforcement layer is calculated by the project similarity calculation method based on user characteristics.The training model generates the recommended candidate set,and finally uses the hidden factor model.The candidate sets are sorted and Top-N recommendations are made.This paper conducts comparative experiments and results analysis on the public data set MovieLense.The results show that the improvement of the RBM model in this paper can improve the recommendation performance.