The Film Website Construction Based On A Collaborative Filtering Recommendation Algorithm

With the rapid development of the Internet, recommended system has beenpenetrated into all parts of the Internet, including film, books, music, news, onlineshopping and so on. As the information boom, recommendation system becomesparticularly important, and the recommendation technology is also gaining theattention of scholars. Collaborative filtering is one of the most important technologyof the recommendation system, and also it is the most widely used and successfultechnology currently.Although the collaborative filtering recommendation technology has achievedgreat success, but the traditional algorithm itself has some problems. These problemsinclude: scalability problem; cold-start problem; scalability problem. Scalabilityproblem directly affects the accuracy of the recommendation system'srecommendation, cold start problems have led to a decline in loyalty for the newusers and the new items can not be recommended. These issues are collaborativefiltering recommendation system must find a way to solve problems.In this paper, these classic problems in collaborative filtering recommendationalgorithm have been made in-depth analysis and exploration. An improvedcalculation formula of similarity was proposed to solve the sparse problem, which isbased a weighting factor, that is the number of the intersection of tow score vectors.In order to solve the cold-start problem, this paper proposes a collaborative filteringrecommendation algorithm based on the item classification to pre-produce the ratings.This method gets the nearest neighbor of the new item by using the properties ofitem itself, and forecast the score of the new item by it's nearest neighbor.Experiments show that the improved algorithm proposed in this paper is superior tothe traditional algorithm in recommendation accuracy. At present the improvedalgorithm has been applied to a movie website that has recommended function. Inaddition to with recommended function, this movie website also have achieved afriend system. By this friend system, you can find and add the users to be your friendswho are similar to you in the taste of movie.Through this research work, sparse and cold-start problems in collaborativefiltering algorithm have been solved to a certain extent.