PERBANDINGAN METODE K-NN DAN BAYES PADA MISSING IMPUTATION

Abstract

The problem of data loss in a dataset is experienced in surveys for data collection which are usually caused by no response from units or items during the survey data collection process. The loss of a data can significantly influence the results of a study. The inaccuracy in choosing a solution to overcome these problems can result in a less than optimal outcome that tends to be biased. Some methods that are widely used to solve these problems are using the K Nearest Neighbor (K-NN) and Naïve Bayes methods, the main purpose of this study is to compare the performance of the two methods. From the results of the K-NN, the results were better, where the Mean Square Error (MSE) is bigger than 1 and MAPE around 10-16%, while Naïve Bayes got MSE values bigger than 1 and MAPE ​​around 26%.