A Study on the Improved Collaborative Filtering Algorithm for Recommender System

The purpose of this study is to suggest an algorithm of a recommender system to increase the customer´s desire of purchasing, by automatically recommending goods transacted on e-commerce to customers. The recommender system has various filtering techniques according to the methods of recommendation. In this study, researchers study collaborative filtering among recommender systems. The accuracy of customer´s preference prediction is compared with the accuracy of customer´s preference prediction of the existing collaborative filtering algorithm, and the suggested new algorithm. At first, the accuracy of a customer´s preference prediction of neighborhood based algorithm as automated collaborative filtering algorithm firstly & correspondence mean algorithm, is compared. It is analyzed by using MovieLens¹ 100K dataset and I Million dataset in order to experiment with the prediction accuracy of the each algorithm. For similarity weight used in both algorithms it is discovered Pearson´s correlation coefficient and vector similarity which are generally used were utilized, and as a result of analysis, we show that the accuracy of the customer´s preference prediction of correspondence mean algorithm is superior. Pearson´s correlation coefficient and vector similarity used in two algorithms are calculated by using the preference rating of two customers´ co-rated movies, and it shows that similarity weight is overestimated, where the number of co-rated movies is small. Therefore, it is intended to increase the accuracy of customer´s preference prediction through expanding the effect of the number of the existing co-rated movies.