Exploring applicability of artificial intelligence and multivariate linear regression model for prediction of trihalomethanes in drinking water

The present study describes the application of artificial intelligence-based modeling approach to predict trihalomethanes in drinking water supplies. The samples were collected from five major water utilities located in five different states across India for two seasons to establish the baseline. Trihalomethane formation was correlated with various operation parameters and exceeded the prescribed guideline value of the World Health Organization and the Bureau of Indian Standard. Chloroform was found to be the most predominant compound > 90% contribution to total trihalomethanes. The seasonal variation assessment revealed that the trihalomethanes level was relatively 1.12 +/- 0.074 times higher in pre-monsoon than post-monsoon. The correlation analysis confirmed, total organic carbon followed by dissolved organic carbon is a major organic precursor responsible for trihalomethane formation. Monitoring these compounds is essential to ensure public safety but cannot be regularly determined due to the involvement of sophisticated instruments and the procedure. The artificial intelligence-based modeling approach could prove to be a good tool for instant prediction of trihalomethanes with better accuracies. An artificial neural network and support vector machine was employed using Python (R) and MATLAB (R) respectively, whereas, for multivariate linear regression, SPSS (R) was used. The value of coefficient and comparison of performance data indicated that artificial neural networks gave the most promising results, followed by support vector machine and multivariate linear regression. The study could prove to be very useful for regulatory agencies to manage and control trihalomethanes' levels in drinking water supplies. [GRAPHICS] .