Design And Implementation Of Personalized Movie Recommendation System

In recent years,Internet technology has developed rapidly,and the services provided by the Internet have increased.What followed is the phenomenon of information overload caused by exponential growth of data.How to filter out useful information for users in a wide variety of Internet data becomes a problem that must be solved.At the same time,with the development of technology,the needs of users are becoming more diverse and personalized.In this case,the recommendation system technology came into being and developed rapidly.Among them,collaborative filtering technology has been widely and successfully applied in various early commercial environments.However,due to the increasing scale of users and items in the recommendation system,the problems which include data sparseness,cold start and long calculation time are becoming more and more prominent.For the problems of data sparsity and cold start,this thesis proposes a hybrid recommendation scheme that uses content-based recommendation to assist collaborative filtering recommendation.At the same time,in order to reduce the rating prediction error to improve the recommendation effect,we will consider the characteristics of user ratings,the number of users who jointly rated the two movies,and user rating time.And we will improve the similarity measure formula of the item-based collaborative filtering algorithm from these three aspects.The core work of the thesis is to improve the recommendation algorithm and design a personalized movie recommendation system by using the improved recommendation algorithm.In terms of recommendation algorithm improvement,this thesis first analyzes the classic collaborative filtering algorithm.By analyzing the characteristics of user-based and item-based collaborative filtering,we have chosen the item-based collaborative filtering algorithm.Then,the similarity measurement method in item-based collaborative filtering is gradually improved.Finally,the one with the smallest recommended error is selected from all the improved similarity measures as the final similarity measurement method used in this thesis.This thesis uses mean absolute error(MAE)to measure the recommended error for each algorithm,because it is widely used and simple to calculate.In order to solve the problems of data sparseness and cold start in collaborative filtering algorithm,this thesis introduces content-based recommendation.The movie content is represented by the movie attribute information.And an attempt is made to comprehensively use item-based collaborative filtering and content-based recommendation to make hybrid recommendation.In the design of the personalized movie recommendation system,this thesis uses the open source framework Grails for web application development.The Grails framework uses the MVC model to make the Web application architecture clear and the modules developed rapidly.The model part of Grails corresponds to the user model,the movie model and the recommendation result model.The control part of Grails corresponds to the back-end recommendation engine service.The view part of Grails corresponds to the front-end user login,browsing,and rating page.Finally,we tested each function module of the system,and they were running normally.