Research On Movie Recommendation Algorithm Based On User Behavior And Feature Tags

With the rapid development of Internet technology,the data problem is becoming increasingly serious.Recommendation algorithm has become the main position to solve the problem of information overload,make efficient use of data and improve user experience.With the booming development of the Internet,each movie platform contains a large number of Movievideos,and thousands of new films are released every year.Although the collaborative filtering technology in the recommendation algorithm has been applied in the movie field,most recommendation algorithms mainly extract the user’s specific behavior information as the basis for recommendation prediction.With the increasing variety of movie scale,the shortcomings of poor user behavior data rules,low algorithm accuracy and low user information utilization are becoming more and more prominent.In view of the low accuracy of the recommendation algorithm and the low utilization rate of user information,this thesis proposes a recommendation algorithm based on user behavior and movie feature tags by using the characteristics of strong regularity and high information accuracy of movie feature tags and scoring data,which introduces the weight of movie feature tag to calculate the prediction score,According to the number of common scoring users and classification,the calculation method of movie similarity is improved.The comparative experiments show that when the MovieLens data set is used,the article coverage rate of the fusion movie label algorithm reaches 87.62%,the accuracy rate of the recommendation result reaches 61.07%,and the recall rate of the recommendation result reaches 36.39%,which effectively improves the performance of the algorithm.The detailed research work of this thesis is as follows:（1）Aiming at the phenomena of poor rules of user behavior data and rapid change of user behavior,in order to effectively meet the evaluation indexes and improve the shortcomings of the current recommendation algorithm,a recommendation algorithm integrating movie feature tags is proposed.Based on the article based collaborative filtering algorithm,the algorithm introduces movie feature tags,and establishes the user movie feature tag matrix,Reconstruct the user’s interest feature model,calculate the recommendation prediction score of the movie through the user’s preference for the feature tag,increase the utilization of the algorithm to the user’s behavior data information,and improve the prediction accuracy of the user’s interest feature model to the user.（2）In order to better optimize the movie similarity set and improve the performance of the recommendation algorithm,thesis improves the movie similarity measurement algorithm from three directions: the internal attribute information of the movie,the actual interval of user scoring operation and the common scoring characteristics of users,and classifies and selects the movie similarity set by setting the similarity threshold according to the classification idea of KNN algorithm,Improve the matching degree between similar movie sets and user interests.（3）The professional movie information data set is obtained from MovieLens website for comparison,and the article coverage,accuracy and recall are selected as the performance measurement indicators of the recommendation algorithm in this thesis.The improved recommendation algorithm is compared with TF-IDF algorithm and article based collaborative filtering algorithm.Through the analysis of the experimental results,it is proved that the improved method in this thesis can improve the recommendation performance.