Comparison of prediction methods for air pollution data in Malaysia and Singapore

Abstract

The process for analyzing and extracting useful information from a large database that employs one or more machine learning techniques is Data Mining. There are many data mining methods that can be used in a variety of data patterns. One of them is prediction modeling. This study compares several data mining performance methods for prediction such as Naïve Bayes, Random Tree, J48, and Rough Set to get the most powerful classifier to extract the knowledge of air pollution data. The parameters being used for observation in the performance of the prediction methods are correctly and incorrectly classified instances, the time taken, and kappa statistic. The experimental result reveals that Rough Set is extremely good for classifying the Air Pollutant Index (API) data from Malaysia and Singapore. Rough Set has the lowest error and the highest performance compared to other methods with the accuracy more than 97%.