Research And Implementation Of Movie Recommendation System Based On Multi-dimensional Data Fusion

The recommendation of current movie and TV website for viewers of relevant movie and video materials is mainly based on the user's recommendation for movie review information and the type of movie.Although the program can recommend some viewers for related videos,the movie types had recommended are relatively simple.At the same time,there is a lack of user feedback on the movie and it is impossible to obtain the user's true taste of the film.The more dimensions of the user's viewing data,the more accurately the user's portrait can be portrayed and the more helpful the user's movie recommendation.Therefore,the multi-dimensional data fusion of the user's viewing perspective is conducive to deepening the user's interest,deepening the user's interest in personalized movies,and improving the system's recommendation quality.In order to solve the above problems,this paper firstly obtains multidimensional data such as facial expression data and viewing behaviors of users,and proposes a movie recommendation model that incorporates multidimensional data.The research work of this thesis is demonstrated in detail by the following points:(1)Personality traits have a great influence on the type of user's preference for film and television works.There is currently related work based on personality traits.We found that when the personality traits are similar,the people with close user backgrounds are more similar because of their exposure to the environment,culture,education,and the life circle.The films recommended by each other are more likely to share the same feelings,and the similarity of group viewing preferences is higher.Therefore,based on the collaborative filtering algorithm based on the personality traits of users,the similarity calculation of user background information is introduced to achieve personalized and accurate recommendation.Experiments show that the method described in this paper has a good effect,which is 8.92% higher than the recommendation error rate when considering only personality traits.(2)User's preference is a time variable that changes with the user's browsing.According to the psychological characteristics,the user's degree of interest in a certain type of movie changes as follows: a certain type of movie that is frequently watched has a high degree of interest;In a movie,the degree of interest rapidly decays over time.In the existing method,the decay rate parameter of interest is fixed,but in this study,it was found that for users with a high degree of preference,because of the frequent viewing of the user,the thinking pattern is formed and the interest rate is attenuated slowly(as long-term memory is formed,Once the memory is formed and it is difficult to forget),the movie that the user does not frequently watch has a faster decay of interest.Based on this,this paper proposes an interest time decay model.based on this,improves the existing recommendation scheme based on time decay.Experiments show that compared with the time-attenuation algorithm based on interestingness,the proposed method proposed in this paper has a certain improvement in quality.(3)The movie comments from user are sparse and random,and they cannot fully reflect true feelings of the user.It is difficult for the recommendation system to obtain sufficient and reliable data to achieve feedback and improve quality.At the same time,due to the lack of quantitative recommendation quality data,users cannot generate sufficient trust in the recommendation system.In order to solve this problem,this paper uses facial expression recognition to capture the actual emotional data of the users in real time and to quantify the user viewing experience.This has a tremendous effect on the feedback and promotion of recommendation quality.At the same time,due to the difficulty of falsifying facial emotions over a long period of time,this method can also effectively prevent the evaluation of water military brushes.(4).In the end,we will combine the recommendation model to complete the design and implementation of the movie recommendation system.The design part mainly includes: the requirement analysis of the recommendation system,the overall architecture design of the recommendation system,the design of the database table,and the design of the user and administrator modules.