Stacking ensemble model with heterogeneous algorithms for the prediction of the water quality index of the Rimac basin

Abstract

Water quality monitoring is essential for the protection of public health and ecosystems. This research used historical data of the physicochemical and microbiological parameters of the Rimac River basin in the city of Lima, Peru, from 2014 to 2021, and proposed a stacking ensemble model with heterogeneous algorithms for the prediction of the water quality index (NSF) in the Rimac River basin/Peru. The results show low values ​​of the mean square error (MSE) and mean absolute error (MAE) of 9.954 and 2.433 respectively. Likewise, a high level of fit with a coefficient of determination of 85.9%. The selection of the prediction model algorithms was based on the detection of stationarity and autocorrelation in the target variable - water quality index. It is concluded that it is necessary to strengthen and use the heterogeneous algorithm to predict the water quality of the Rimac basin. It was developed in a Google Colab environment and Python programming language