Design And Implementation Of A Diversified Movie Recommendation System

With the rapid development of the Internet and the popularization of mobile terminals,a big volume of information has brought about the problem of "information overload".A lot of meaningless information has caused great trouble for people to select valuable information.In view of this situation,the recommender systems emerge at the right time.Through analyzing the records of a user’s historical behaviors,a recommender system can further master the user’s personal preferences and provide the information needed.Collaborative Filtering(CF)has become one of the most commonly used technologies in current recommender systems,whose biggest feature is "domain independence".It can make recommendations for target users based on the similarity among users.However,the collaborative filtering technology still has some defects in the process of practical application introduced as follows:(1)the traditional collaborative filtering recommendation methods has pretty low efficiency.Especially in the era of big data,the decisionmaking data used for providing recommendation changes constantly as time goes on.Consequently,the calculation process will cost a lot of time,which may make it impossible for them to meet the quick response requirements of users;(2)most of the traditional collaborative filtering methods pay more attention to the accuracy of the recommendation results,and prefer that the items in the recommended lists are independent among each other.Thus,there might exist redundant or duplicated items in a list,which may not cover a wide range of user’s interests and further decreases the users’ satisfactions.Pointing at the problems mentioned above,in this article,a detailed study on the improvements of the traditional collaborative filtering technology is implemented,and an improved method is applied to develop a diversified movie recommender system on this basis.The specific research work are as follows:(1)Firstly,in this article,the development background,application situation and collaborative methods of the recommender system are introduced in detail,and some defects of the collaborative filtering methods are analyzed.(2)Then,the locality sensitive hashing algorithm is introduced into the recommendation problem to solve the problems of low efficiency and redundant results of recommendations and further improve the recommendation effects.This method can build item index offline and as a result,the dimensionality of high-dimensional data can be reduced.Afterwards,according to the item indices,we quickly search for the similar items of a target user’s history records,which can improve the efficiency of recommendation greatly.In the process of constructing the recommended list,we use rating diversity to achieve the quadratic optimization,which can eliminate the similar items in the recommended list.It can ensure the items in the recommended list are dissimilar to each other with a high probability and improve users’ satisfactions by providing users with a diversified recommended list.To verify the effectiveness of the improved method,in this article,the dataset of MovieLens is selected and comparison tests are conducted to measure the commonly-used indicators of the recommendation system.In conclusion,the experimental results proved that the proposed method is feasible.(3)Last,the above improved method is applied to the movie recommender system,and the diversified movie recommender system is designed and implemented.First of all,we analyze the feasibility,functional requirements and non-functional requirements of the system comprehensively.Then,the overall design of the system is completed.Moreover,we develop the functional modules of the movie recommender system with Java language in the Eclipse development environment.