Design And Implementation Of A Film Recommendation System Based On Mahout

The popularity of computers and the rapid development of the Internet has brought the problem of information overload,people are troubled by how to get the information they want,and recommended the birth of the system effectively solve this problem,it can not only help users find Information on their own valuable,but also allows information to show interested in it before the user.The main part of the system is the recommendation algorithms.The most widely used and most mature one is collaborative filtering,which makes use of similar users or similar items for recommendation by sharing the historical experience in the user community.A recommendation system typically relies on a website or application.Mahout-based movie recommendation system is divided into two parts,offline recommendation engine needs to achieve four typical recommended scenes,including popular movie recommendation,the latest movie recommendation,guess you like and related movie recommendations,and online display part of the MVC framework Development,the need to achieve the basic functions of a website.From the business logic division,the overall structure of the system is divided into five layers,functional modules,including off-line functional modules and online functional modules.The core algorithm design part introduces the relevant implementation class of the recommendation engine part of Mahout,and uses the open source data set to carry on the simulation experiment.Through analyzing and evaluating the experimental results,we choose the appropriate collaborative filtering algorithm for the personalized recommendation requirement.Similarity calculation method.Then,according to the requirement analysis and the system design,the database is designed with the E-R chart.Then the Hadoop,Mahout and Spring MVC are used to realize the online function module of the recommendation engine algorithm and recommendation system.Finally,test cases are designed for each functional module,and functional tests and non-functional tests are carried out to verify whether the system is in line with expectations.The recommendation engine runs the recommendation algorithm on a regular basis offline,processes the user history score data,and updates the results to the database.Although the engine can not handle the newly generated scoring data in real time,this mechanism improves the loading speed of recommendation results and reduces the waiting time of users.