A Preliminary Study On Anime Recommendation:based On Audience Clustering In Trust Relationship

In recent years,a growing open market and explosive vigour have been shown in the anime industry.Meanwhile,personalized recommendation research has also attracted much attention in the era of big data.This paper is dedicated to exploring the problems of rating prediction in the application scenarios of anime recommendation.Based on the trust relationship of anime audience,clustering and collaborative filtering technologies are integrated,in order to improve the accuracy of traditional methods in anime rating prediction problems and improve its inherent sparsity defect.According to the characteristics of anime audience,this paper defines the trust relationship between users,and pre-clusters the animes to extract users’ characteristics of their interests in various animes.This paper proposes a distance measurement strategy for mixed-feature data,which uses Gower distance,word frequency matrix and Jaccard distance,and reasonably sets the weight coefficients.This strategy is helpful to obtaining overall distance and furthermore deriving the significantly explanatory and practical clustering results.This paper proposes an improved user-based collaborative filtering method with clustering and group-individual trust relationship model(K-GITUBCF).Through the filtering effect of group trust between groups and the correction effect of individual trust within group,small-scale search for neighbours and improvement of similarity calculation are realized.It is applied to the real datasets obtained by MyAnimeList.net that contain users,animes and ratings information.Supported by experimental results,K-GI-TUBCF shows a better performance on anime rating prediction,than other related existing methods.From the perspective of media platforms in the middle of the anime industry chain,this paper combines mainstream popular technology of personalized recommendation to construct a predictive recommendation method suitable for the anime field,thereby contributing some commercial value to the actual business development.