Machine Learning-Based Pollution Risk Classification in Industrial Wastewater Treatment
Mustafa Ozan Ozkok
Source abstract
Industrial wastewater contains various pollutants that can impact water quality and the environment. It is important to identify pollution risk conditions in order to monitor wastewater treatment processes effectively. This study compares four machine learning algorithms — Random Forest, Support Vector Machine (SVM), XGBoost and LightGBM — for classifying pollution risk. The Melbourne Eastern Industrial Wastewater Plant dataset was used for this study, containing 600 observations and 14 physicochemical and operational variables. The dataset includes two predefined classes: 'Normal' and 'High Risk'. The models were evaluated using stratified five-fold cross-validation. Accuracy, precision, recall, F1-score and ROC-AUC were used to compare their performance. XGBoost achieved the highest mean accuracy of 93.5%, precision of 90.32%, recall of 91.43%, F1 score of 90.72% and ROC-AUC of 0.9803. LightGBM achieved an accuracy of 93%, followed by random forest with 91.67% and SVM with 90.33%. SHAP analysis was used to examine the contribution of input variables to XGBoost predictions. According to the mean absolute SHAP values, COD and BOD5 were the most influential variables. The results showed that the models could distinguish between the predefined pollution risk categories. However, the risk classes were derived from a composite pollution index, and further validation using independent datasets is needed.
Evidence graph
No public relationships recorded yet.
Integrity note: This page is a factual metadata record created by deterministic ingestion. It is not a claim that the work moves a mathematical frontier or has been independently verified.