Forecasting Total Dissolved Solids to Predict Ground Water Quality Using Machine Learning Techniques

Water is vital to the existence of every living thing on Earth. Nonetheless, a variety of pollutants including industrial, agricultural and inorganic waste have contaminated drinking water. This study aids in the prediction of quality of ground water for drinking purpose. To anticipate good WQI(Water Quality In-dex), it is necessary to include all crucial parameters rather than ignoring them. Regression Techniques are used to anticipate the parameters value based on the historical data. One of those parameter in this study is Total Dissolved Solids(TDS). Regression models such as Random Forest, Decision Tree, XGBoost, Lasso and Ridge are used to forecast the value of the TDS. The Lasso Regressor outperformed the other regressors with an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{R}^{2}$</tex> score of 0.95%. The values of the Water Quality Index (WQI) were determined using the Weighted Arithmetic Water Quality Index (WAWQI). Upon estimating WQI the values have been segregated according to their respective Water Quality Classes (WQC) like Excellent, Good, Poor, Very poor and Unsuitable categories by giving specific range from 0 to 100 with intervals of 25 using the classifiers such as Decision Trees, Gradient Boosting, Random Forest, KNN, XGBoost and CatBoost. XGBoost performed better than the others by producing an accuracy of 0.96%. Evaluation of classifier were carried out using the following performance metrics: F1 Score, Accuracy, Precision, and Recall.