Predictive Modelling of Photocatalytic Wastewater Treatment Using Random Forest and XGBoost on Mn-Doped BiFeO₃

Photocatalytic wastewater treatment must have precise prediction models to enhance performance and minimise experimental endeavours. Current models, including basic decision trees and gradient boosting, frequently encounter difficulties in managing non-linear correlations in photocatalytic data, resulting in decreased accuracy and dependability. This study presents a hybrid predictive model that integrates Random Forest (RF) with XGBoost, employing an interpolation approach to improve performance. The hybrid methodology decreases the overfitting associated with Random Forest and the sensitivity inherent in XGBoost, resulting in an accuracy of 95%, an F1 score of 92%, and a recall of 91%. This approach has enhanced prediction skills relative to conventional models, rendering it particularly beneficial for complex wastewater treatment datasets. The suggested technique significantly enhances prediction robustness and provides a scalable solution for environmental applications.