An Optimized Approach for Predicting Water Quality Features Based on Machine Learning

Traditionally, water quality is assessed using costly laboratory and statistical methods, rendering real-time monitoring useless. Poor requires a more practical cost-effective solution. The machine learning classification approach appears promising for rapid detection prediction of quality. Machine has been used successfully to predict However, research on index (WQI) generally lacking. Therefore, this aims identify the important features WQI, which necessitated numerous indicators. This study develops four models (Artificial Neural Network, Support Vector Machine, Random Forest, Naïve Bayes) based WQI chemical parameters. Langat Basin in Selangor dataset from Department Environment Malaysia trains validates each model. Several data preprocessing tasks such as cleaning feature selection have conducted raw ensure training data. performance these algorithms further rectified selected set by several strategies information gain, correlation, symmetrical uncertainty. Each classifier then optimized different tuning parameters achieve optimum values before comparing output three classifiers against other. observational results shown that Forest with parameter gain method achieved highest performance. experimental show are relevant predicting than other variables. Consequently, result shows oxygen (DO) biochemical demand (BOD) WQI. proposed model reasonable accuracy minimal parameters, indicating it could be systems.