Entropy based Fuzzy C-Mean for Item-based Collaborative Filtering
Kiatichai Treerattanapitak, Chuleerat Jaruskulchai · 2009
Collaborative Filtering (CF) is a method behind the successful of recommendation system by attempting to predict user interested from peer's opinions and recommend similar items that match user's interested. A Challenge for collaborative filtering is data characteristics. It's always contains a lot of missing values either by gotten number of rating from user is very low or new items add to the system. Consequently, accurate prediction is difficult due to there is not much data can be used to train the system. In addition, the computational cost of CF becomes highly expensive on very large data size therefore users expect real-time recommendation but CF could fail to operate because of data size. Many approaches tried to overcome these issues, fuzzy C-mean for Item-based approach (IFCM) is one of efficient and easy to reduce impact of scalability by performing clustering on the item side. In this paper we employ entropy based to IFCM. Our result shows that IFCM+entropy improve MAE by 3.2% and 13.4% on single rate items compare to IFCM.