Enhancing the performance of evaluation in accurate prediction of water quality with two layered model using regression algorithms.

River water pollution is a critical issue of enormous importance. India faces significant challenges in maintaining the quality of its water resources as a country with a large population and rapid industrialization. River water pollution has far-reaching consequences that go beyond environmental concerns and directly affect the health and well- being of millions of people. India heavily relies on its river water bodies for a variety of purposes, including drinking water, agriculture, industrial processes, and ecosystem maintenance. The findings suggest that, while a single approach may be accurate at certain times, its accuracy may vary at other times. This issue was observed in both machine learning such as Linear Regression (LR), Partial Least Square Regression (PLSR) and Random Forest regression (RFR) and statistical models, including the models under consideration in this study. Despite the promising results of a two-layered stacking model in water pollution research, obstacles remain. The two-layered stacking model (RFR) predicts Water Quality Index (WQI) with greater accuracy. In terms of accuracy and predictive ability for WQI, the performance metrics show that this model outperforms the base regression models (LR, PLSR, and RFR). We implemented this model stacking approach with Indian rivers and discovered that it forecasts with high accuracy among all five base models by cross-validation of their performances. The accuracy ranking of stacking model remained consistent at 99.15%. As a result, it is recommended that a combination of water quality parameters be used to evaluate the performance of regression algorithms for WQI prediction. [ABSTRACT FROM AUTHOR]