Research And Implementation Of Personalized Movie Recommendation System

With of the constantly increasing of network resources and the rapiddevelopment of video sites, the information retrieval service basing on the traditionalmode has been very difficult to meet the needs of users. Therefore, the personalizedrecommendation service are more broadly concerned,it can excavate userinformation, movie information, the user’s operating log, and the hidden correlationbetween the data, recommending to the user for the gained movies and televisionwhich the user may be interested in.This paper discuss and analysis the issue about the data mining technology andits classic association rules data mining algorithm,such as （Apriori algorithm、FP-Growth algorithm）. According to Issues of the apriori algorithm that exist in themining process,such as, generating frequent itemsets need repeatedly scan thedatabase, having a large number of candidate sets in the middle processing,so theimprovement of the apriori algorithm is proposed in this paper, and the text name it asfpmdf algorithm.The basic idea of the fpmdf algorithm is using dynamic function for frequentpattern mining. It is based on the transaction ID of the transaction database pair toproduce a candidate set, Every iteration is based on the candidate set which isproduced by the previous iteration, so no need to repeatedly scan the transactiondatabase, reducing the running time of the algorithm, and the experiments provedthe advantages of the algorithm in the time and space. In addition,By comparing theoperating efficiency in the Apriori algorithm、FP-Growth algorithm and the improvedalgorithm-fpmdf, Experiments are performed to verify the feasibility of thealgorithm with existing experimental data sets.Finally, this paper using the association rule mining improved algorithmdesigned and realized an personalized movie recommendation system. Personalizedmovie recommendation system generally recommended the itemsets that includespopular videos, new film recommended, recommended related videos.This thesis is focused on how to mine data from a large number of historical data in the watching,finding out the correlation association rules between the different users and movies aswell as movies and movies. Extracted rules and as a reference for the movierecommendation system providing the active recommendation service, providing thepersonalized,related movies recommendation for users, and designed an experimentused to verify the recommendation system of satisfaction.