A Hybrid Approach Based on Principal Component Analysis and Artificial Neural Network for Modeling River Water Quality of Prayagraj Region – An Important Pilgrimage Site of India

The present study investigates the quality of river water of Prayagraj region, which is a prominent holy pilgrimage site of India. A total of 100 river water samples were collected from five selected sites during January 2020–November 2021 and examined for 12 physicochemical parameters. The mean values of turbidity (6 ± 3.65 NTU), total hardness (220 ± 25.69 mg L −1 ), and dissolved oxygen (8.2 ± 2.74 mg L −1 ) exceeded the BIS permissible limits. The physicochemical parameter dataset was employed to evaluate the weighted arithmetic water quality index (WQI). River water of the region was found to have an overall WQI of 66.39 ± 23.35, suggesting that it was unfit for human consumption, although it may be utilized for agricultural and industrial applications. Recognizing the importance of computational modeling for rapid WQI evaluation, 18 machine learning algorithms were investigated, which revealed PCA‐ANN (principal component analysis‐artificial neural network) as the most appropriate model for WQI prediction. PCA reduced the dimensionality of water quality data by identifying 7 PCs that explained >90% variance. The model ANN‐A directly incorporated all 12 physicochemical parameters as input, achieving an exceptionally high accuracy ( R 2 = 0.998, RMSE = 0.829, A f = 1.088, MAPE = 0.847%). The model ANN‐B utilized PC scores of only 7 significant components and could attain a commendable accuracy ( R 2 = 0.950, RMSE = 5.193, A f = 1.289, MAPE = 7.854%). Consequently, PCA‐ANN may serve as a significant and reliable tool for policymakers for modeling WQI, since it not only reduced the data load but also generated a decent prediction accuracy.