Biochemical and human-made pollution of river water can introduce water discharges to potentially toxic diseases. Forecasting is an alternative approach for watershed management that identifies numerous restrictions related to conventional traditional approaches to evaluating the quality of water. Artificial intelligence algorithms and machine learning (ML) approaches have been broadly and successfully used to classify the decision-making issues such as more inconsistent behavior through one year or river to the next. As a result, the perfect approach for machine learning varied based on the river and the year, helping to make methodology challenging. This work compares performances of the different ML approaches referred to as a model and their corresponding output of four widely used machine learning models (multiple linear regression (MLR), partial least square regression (PLSR), random forest (RF), and ensemble random forest (ERF)) were compared and identifies the more accurate final prediction. Applying this model to various rivers situated all over India, we demonstrate that the Linear Regression algorithm was able to produce reliably good predictions of river water quality compared to all of the other approaches. Performances of the above mentioned models were tested by calculating the value of Root Mean Square Error (RMSE). The accuracy levels consistently stayed every year, with accuracy of 96.06%, 99.99%, 94.49%, and 93.6% at the ensemble random forest, multiple linear regressions, partial least square, and random forest regression respectively. The RMSE value of these models was 53.71, 1.14, 68.5, and 50.67 respectively. This work evaluates the significance of the linear regression approach in forecasting the water quality of Indian Rivers and resolving other demanding ecological issues.