Research On Hybrid Recommendation Method Based On Collaborative Filtering

Recommendation system is an information filtering tool that can help users find the products or services they need from a large amount of information.Collaborative filtering(CF)technology is an important technology in recommendation system,which can provide users with personalized recommendations.However,the user-item rating matrix is sparse and a large number of ratings are missing,which results in only a few ratings can be used to predict unknown ratings.In order to solve the problem of data sparsity,the existing collaborative filtering technologies need to be studied and improved.On the basis of consulting relevant domestic and foreign literature,the thesis analyzes and summarizes the relevant theories and technologies of recommendation system,and carries out the following research work:(1)Aiming at the problems of low prediction accuracy and poor effect of traditional CF recommendation algorithms in the case of sparse data,an improved hybrid recommendation model based on collaborative filtering is proposed.When calculating the similarity between users or items,the model adopts cosine similarity and Pearson correlation coefficient.It combines user-based collaborative filtering,item-based collaborative filtering and linear regression model,which can accurately predict the relationship between unknown users and items.Through experiments,the prediction performance of the improved hybrid recommendation model based on collaborative filtering is compared with the existing models such as RSVD and LDA.The results show that the hybrid model has higher recommendation accuracy and less computation than other models.(2)Aiming at the problems of lack of scoring information and low coverage in the case of sparse data in the traditional CF recommendation algorithms,a hybrid multi-criteria model for personalized recommendation is proposed.The model uses the concepts of multi-criteria scoring,implicit similarity,similarity transitivity and global reputation,which can better establish the relationship between users or items.It combines CF based on multi-criteria users with CF based on multi-criteria items.When the data is sparse,it adopts the switching hybrid strategy to switch between the user-based and item-based modules.It solves the problem of data sparsity in the CF based recommendation system,improves the prediction accuracy and coverage,and can be applied in a variety of recommendation fields.Through experiments,the hybrid multi-criteria model is compared with the user-based and item-based single criteria CF models and the user-based and item-based multi-criteria CF models.The results show that in the case of sparse data,the model has obvious advantages in prediction accuracy and coverage compared with the benchmark CF recommendation models.