Cross Domain Recommendation Method Based On Similarity Self-Learning

With the advent of the big data era,users are facing the problem of information overload.In order to meet the individual needs of users,recommender system came into being.The recommendation system is to take the initiative for users from a wide range of information in the ocean to find the user favorite information products of the times.Among them,the cross-domain recommendation is aiming at the problem of sparse data and cold-start users in the single-domain recommendation,and a method for recommending a plurality of auxiliary domains in combination is proposed.At present,the research recommended in the cross-domain mainly focuses on how to improve the accuracy of the recommendation through various methods or considering more comprehensive factors.In cross-domain recommendation,On the one hand,in view of the randomness of users' shopping scores,Based on the user's existing rating data to calculate the user rating reliability,and then through the reliability of the score segmentation thresholds,the number of scores below the threshold score cleared cross-domain recommendation pretreatment to reduce the auxiliary domain random score interference;on the other hand,In the cross-domain recommendation model based on user similarity self-learning,The user similarity calculation has been more comprehensive consideration,adding common items,rating time,user rating three factors comprehensive calculation of user similarity.At the same time,when calculating the similarity of shared users in the auxiliary domain and the target domain,a selection strategy based on the target domain and the auxiliary domain is proposed,which in theory makes the recommendation method more realistic.Finally,the cross-domain recommendation method based on reliability of scores and the cross-domain recommendation method based on self-learning of user similarity are combined to be recommended.This thesis is validated and compared on the Amazon dataset,Experimental results show that cross-domain personalized recommendation method based on reliability of ratings can improve the accuracy of recommendation,As well as the improved user similarity method based on cross-domain recommendation method of self-learning of user similarity,the selection strategy of similarity value of shared users respectively improves the accuracy of recommendation.Finally,the cross-domain recommendation method based on reliability of scores and the cross-domain recommendation method based on user similarity self-learning can effectively improve the recommendation accuracy.