Machine Learning by Data Mining REPTree and M5P for Predicating Novel Information for PM10

Abstract

We examined data mining as a technique to extract knowledge from databases to predict PM10 concentration related to meteorological parameters. The purpose of this paper is to compare two types of machine learning using data mining decision tree algorithms—Reduced Error Pruning Tree (REPTree) and Divide-and-Conquer M5P—to predict Particulate Matter 10 (PM₁₀) concentration based on meteorological parameters. The results of the analysis showed that the M5P tree gave a higher correlation compared with REPTree, as well as lower errors and a higher number of rules. The elapsed processing time for REPTree was less than the processing time for M5P. Both of these trees indicated that humidity absorbs PM10. The paper recommends REPTree and M5P for predicting PM10 and other pollutant gases.