Research On Movie Recommendation Algorithms Sensitive To Emotional Factors

Context-aware recommendation system is one of the current research hotspots in the recommendation field.Its emergence breaks through the limitation that traditional recommendation algorithms can only rely on the binary relationship between users and items,and takes into account the context of users and items,such as time,location,Seasons,weather,surrounding people,emotions,etc.,improve the recommendation performance of the recommendation system based on the ternary relationship of usersituation-items,and can be widely used in movie recommendation,live broadcast recommendation,product recommendation and many other scenarios.Movies are a strong emotional carrier,so in the field of movie recommendation,especially emotional information,is particularly special and important.However,existing movie context-aware recommendation systems often have problems such as ambiguous contextual meaning,insufficient use of contextual information,and high computational complexity.The three-dimensional matrix of user-context-item will also cause the problem of increased data sparseness.A current research point of view is that statisticalbased algorithms have approached the bottleneck,with extremely poor interpretability and difficult to improve qualitatively.It is necessary to introduce other cross-disciplinary achievements to help continue to improve the recommendation effect in a specific field.In order to solve the above-mentioned problems,the theory of emotion regulation in China and the West was studied,and a TCM psychology theory was selected,which was introduced into the design of recommendation algorithm.A movie "emotion-oriented model" and two movie recommendation algorithms that are sensitive to emotion are proposed,which not only Significantly reduces the computational complexity,but also makes the recommendation results easy to interpret: First,presume five types of"emotional orientation" based on the TCM psychology theory",and then use a large number of emotional tags marked by movie viewers for the movie,and refer to the idea of k-means clustering algorithm and TF-IDF algorithm to establish " Emotion-oriented model" for the movie;Secondly,the SVD++ and ICAMF algorithms were improved based on the "emotion-oriented model",and designs two movie recommendation algorithms that are sensitive to emotion: Emotion_SVD++ and Emotion_ICAMF.Finally,use the real data set collected by Douban.com and the public data set of LDOS-Co Mo Da for testing and comparison.The results show that after Introducing the emotional regulation theory,the RMSE and MAE predicted by the new algorithm are reduced compared with RSVD,Bias_SVD,SVD++,and ICAMF-I,and algorithm convergence time is significantly shortened than the best reference algorithms: ICAMF-? and ICAMF-?.