Predicting Kereh River's Water Quality: A comparative study of machine learning models

Abstract

This study introduces a machine learning-based approach to forecast the water quality of the Kereh River and categorize it into 'polluted' or 'slightly polluted' classifications. This work employed three machine learning algorithms: decision tree, random forests (RF), and boosted regression tree, leveraging data spanning from 2010 to 2019. Through comparative analysis, the RF model emerged as the most efficient, boasting an accuracy of 97.30%, sensitivity of 100.00%, specificity of 94.74%, and precision of 95.00%. Notably, the RF model identified dissolved oxygen (DO) as the paramount variable influencing water quality predictions.