Determining the Effectiveness of Learning Techniques for dust Classification using data mining Techniques

Abstract

This research aims to create a model for dust efficiency in the air to compare the efficiency of models for dust forecasting using 4 data mining techniques, consisting of Naïve Bayes, Random Forest, K-Nearest Neighbor, and Decision Tree. The data from the Office of Natural Resources and Environmental Control 9, Udon Thani Province, were used to create a model set and a model test set, which were data from a review of a retrospective air quality data set of 724 items with 7 characteristics. The accuracy was then calculated by The research results can be summarized as follows: Random Forest Method The most efficient in forecasting with an accuracy of 98.03%, the decision tree method has an accuracy of 91.83%, the Naif Bayes method has an accuracy of 88.35%, the nearest neighbor K-NN method has an accuracy of 85.03%, the Neural Net method has an accuracy of 91.06%, and the SVM method has an accuracy of 85.49%,respectively. It was found that the random forest method has the most efficient in modeling compared to the combined comparison methods because it is a non-distribution or non-parametric method that does not depend on the probability distribution assumption. It is suitable to use the model for forecasting as a guideline to support environmental decision-making and pollution control in air quality forecasting.