Improving Water Quality Index Prediction Using Regression Learning Models.

Area of Science:

• Environmental Science
• Data Science
• Water Resource Management

Background:

• Rivers are vital freshwater sources threatened by pollution from economic activities.
• Monitoring river water quality involves parameters like pH, dissolved oxygen, and total suspended solids.
• Predicting water quality trends aids in pollution prevention and restoration effectiveness assessment.

Purpose of the Study:

• To apply machine learning regression techniques for predicting river water quality index.
• To compare the performance of eight different regression algorithms.
• To evaluate the impact of derived features on model accuracy.

Main Methods:

• Utilized historical data from Indian rivers, focusing on six key water parameters.
• Derived twelve additional features from the original six parameters.
• Applied eight machine learning regression algorithms: decision tree, linear regression, ridge, Lasso, support vector regression, random forest, extra tree, and artificial neural network.

Main Results:

• Derived water quality rating scale features significantly improved regression model development.
• Linear regression and ridge algorithms demonstrated the best performance among the tested models.
• Achieved a best mean square error of 0 and a correlation coefficient of 1.

Conclusions:

• Machine learning regression is effective for river water quality index prediction.
• Feature engineering, specifically derived rating scales, enhances predictive model performance.
• Linear regression and ridge models offer a robust approach for water quality forecasting.