The Characteristics of Response and Top-N Accuracy in Collaborative Filtering of Recommender System

This research suggests a way to predict recommending accuracy of Top-N in advance when recommending Top-N to customer’s in recommender system using collaborative filtering algorithm. Response distribution has to do with the recommending accuracy, and the recommending accuracy can be predicted by the statistic of response distribution. The prediction of recommending accuracy can be a useful method in deciding the number of recommending lists. The recommending accuracy can increase the customer’s satisfaction and also reliance of it. Through the results of this research, customer’s satisfaction is expected to be increased. To increase the accuracy of estimation value in the recommender system, it is important to decrease the error of prediction value. However, when predicting in reality, it is recommended by the estimation value and recommendation lists are made by the estimation value. Thus, making the list of Top-N is very important.