Using machine learning and linear regression to forecast the water quality in Al-Sain lake

Abstract

Effective management of the quantity and quality of water requires accurate assessment and determination of the pollution levels of surface and groundwater. The goal of this study is to assess the effectiveness of multiple linear regression (MLR) and 19 machine learning (ML) models, which utilize various algorithms such as regression, boosting, and decision tree. Among of these models are linear regression (Lr), least angle regression (Lar), Bayesian ridge chain (Br), ridge regression (Ridge), k-nearest neighbors regression (K-nn), extra tree regression (Et), extreme gradient boosting (XGBoost), etc. By employing these models, the study aims to accurately predict the surface water quality of Al-Sain lake in Latakia city.

To define the water quality index (WQI), data from the drinking water lake intake for the years 2021-2022 were analyzed. The effectiveness of the multiple linear regression (MLR) and machine learning (ML) models were assessed using statistical tools such as the coefficient of determination (R2) and the root mean square error (RMSE) to gauge their accuracy.

The results indicated that the multiple linear regression model (MLR) and 3 of the machine learning (ML) models, including linear regression (Lr), least angle regression (Lar), and Bayesian ridge chain (Br), performed extremely in predicting the (WQI) index with high accuracy (R² = 0.99, RMSE = 0.15) for the (MLR) model, and high accuracy (R² = 1.0, RMSE ~= 0.0) for the three aforementioned machine learning (ML) models. The results support the use of multiple linear regression models and machine learning models in predicting the water quality index (WQI) with very high accuracy, which will contribute to improving of water quality management.