Research And Implementation Of Hybrid Recommendation Algorithm Based On User Interest And Topic Model

With the rapid growth of Internet information,people cannot obtain satisfactory information from a large amount of information,leading to the problem of information overload.Personalized recommendation algorithm has become the main technology to solve information overload.At present,domestic and foreign scholars have conducted related research on the recommendation algorithm.Data sparsity problem,cold start problem and user interest acquisition problem are all important factors affecting the recommendation effect.In view of the above problems,this paper studies the recommendation algorithm based on user interest and topic model.The main research contents of this article are as follows:Based on the user interest recommendation,a collaborative filtering recommendation algorithm based on the mixed similarity is proposed for the similarity measurement problem,aiming to find nearest neighbors with similar interest to active users.The algorithm adopts the method of segmented similarity measure.It not only can analyze the global impact of user's rating behavior,but also can make full use of user attribute similarity and item similarity to improve the accuracy of user's similarity.The experimental results show that compared with other algorithms,the proposed algorithm mitigates the problems of cold start and data sparsity,and improves the accuracy of the prediction and the recommended performance.Based on the topic model recommendation,a collaborative filtering recommendation algorithm based on the improved topic model is proposed to accurately obtain user preferences for the integration of time and score information in the model.The algorithm generates time function by fitting Ebbinghaus forgetting curve and combining time window,and then changes the data weight,finds the changing law of user interest,calculates the distribution probability of the item under specific interest according to the feedback of user's scoring behavior,and mines the group user.The effect of the behavior on active users.Experimental results show that the improved recommendation algorithm can more accurately capture user interest and improve recommendation performance.A personalized movie recommendation system is designed and implemented.According to the user's history scoring behavior in the system,two recommendation algorithms are mixed in a linear weighted manner to generate the final result,and then recommended the result to user.The system test verifies that the system can be personalized according to different users,and has the advantages of easy operation and good performance.