<![CDATA[Recommender systems are basically developed to help internet users to find information such as movies, books, news or other informations that probably prefered personally. In this study we constructed a movie recommender system using cluster-based smoothing method of collaborative filtering that can provide movie predictions for a user with good accuracy. The implementation of clustering and smoothing of this method is expected to address the issue of data gap rating (sparsity) and an increasing number of users in large numbers (scalability) that often occurs in collaborative filtering. The test of this method is using three parameters, namely the number of clusters, the number of k-neighbors, and the level of sparsity. In the test from each parameter is used dataset rating as much as 1063 instance that has been rated by 40 users, then the data is divided into 2 parts, with 80% as data train and 20% as data test. Test results showed that the number of clusters = 3 and number of k = 50% produced the lowest accuracy value is 0.6713 and the system is capable to handling level of data sparsity from up to 70% with MAE = 0.8361.Keywords: Cluster-based Smoothing, Collaborative Filtering, movie, recommender system, Scalability, Sparsity]]>
Copyrights © 2012
