MACHINE LEARNING-ASSISTED PREDICTION OF BACTERIAL INACTIVATION BY COBALT FERRITE-ACTIVATED HYDROGEN PEROXIDE

Abstract

Hydrogen peroxide activated on cobalt ferrite is a promising non-chlorine option for continuous-flow water disinfection, but bacterial survival depends nonlinearly on oxidant dose, initial load, and contact time. This study compared Weibull and Gompertz equations with ten supervised machine-learning algorithms using 208 individual Escherichia coli observations collected at 3, 6, and 10 mM H₂O₂. Gradient boosting and random forest performed best; their averaged model reached a group-based R² of 0.963. Independent experiments at 4 mM H₂O₂ confirmed good interpolation, supporting dose-contact-time selection within the tested range.