Imputation missing value to overcome sparsity problems

Abstract

Collaborative filtering (CF) is a method to be used in recommendation systems. CF works by analyzing rating data patterns from previous users to produce recommendations according to their interests. However, it faces a crucial problem, sparsity, a condition where a lot of data is empty, which will affect the quality of the recommendations produced. To state this problem, the purpose of this study is to input methods including mean, min, max, and k-nearest neighbor imputation (KNNI). The steps taken include imputation of empty data, followed by similarity calculations using the cosin similarity method, and evaluation using root mean square error (RMSE). The experimental result shows that the mean method is excellent with an average similarity value of 0.99 and an RMSE value of 0.98.