Second-order based ensemble machine learning technique for modelling river water biological oxygen demand (BOD): Insights into improved learning

Abstract

Generally, water bodies are composed of a small amount of organic matter that affects their quality for both domestic and industrial applications. Therefore, the current study involves modeling the Biological Oxygen Demand (BOD) using various physicochemical parameters. The study involves implementing different stand-alone, first-order, and second-order ensemble paradigms. Based on the quantitative and visualized results obtained from the current research, both stand-alone and first-order ensemble paradigms failed to model the BOD with reliable performance. Therefore, the current study proposed the first application of second-order ensemble machine for modelling BOD in the literature. The comparative performance of second-order ensemble paradigms indicates the strong ability of non-linear paradigms over linear methods. Whereby, 2′-AE with DC = 0.992, R = 0.996, RMSE = 0.136 and 2′-NNE with DC = 0.932, R = 0.966, RMSE = 0.248 depicts satisfactory and reliable performance in both the training and testing phases in modelling the BOD.

Therefore, the proposed technique can serve as a satisfactory approach for various hydrologists, policymakers, and decision-makers in the treatment of water from different water treatment plants.