Leveraging Supervised Machine Learning Algorithms in Predicting Contaminants Removal From Novel Cascade Aerating Trickling Filter

Abstract

This research evaluated machine learning (ML) algorithms to predict the efficiency of a cascade aerating trickling filter (CATF) in removing physicochemical characteristics such as suspended solids (SS), biological oxygen demand (BOD) and chemical oxygen demand (COD). The pre‐evaluation of characteristics' data demonstrated that CATF is effective in removing TSS, BOD and COD. Four ML algorithms, that is, random forest (RF), Gaussian process regression (GPR), extreme gradient boosting (XGB) and light gradient boosting machine (LightGBM), were assessed for physicochemical prediction of the CATF. The appraisal of the ML algorithms revealed that RF outperformed the others in predicting 81% SS and 84% COD removal. Nevertheless, GPR was found to be superior, with 66% accuracy in detecting CATF's BOD performance. Ensemble ML algorithms, such as RF and GPR, are recommended as effective solutions for optimizing the performance of wastewater treatment plants.