Water quality forecasting using a Novel Hybrid DNN-MBGD optimization and WAWQI technique for assessment of surface water quality index in 10 districts of Uttar Pradesh

Abstract

Chemicals from farm fields, salts from the road, pollutants from the atmospheric circulation, garbage from domestic and industrial wastes, hazardous wastes and animal wastes can lead to water pollution. Chemicals, bacteria and viruses from the septic system can also infect water contaminated with synthetic materials such as gasoline. This polluting water affects human health and is unsuitable for drinking, so water quality is important. This study estimates the water quality index by the Weighted Arithmetic Water Quality Index Method (WAWQI) using 20 water quality parameters in Uttar Pradesh for 10 districts in one year (January 2019–January 2020). Also, a novel hybrid Deep learning Neural Network-Mini Batch gradient descent optimization (DNN-MBGD) is used for water quality prediction. Based on the results, water from the experimental sites is unsuitable for drinking and other purposes like domestic and irrigation. The WQI value for these 10 sites is much greater than 100. After WQI calculation, the relationship between two water quality parameters are determined by a correlation matrix. Based on the feature importance score, input features were selected, and the performance of K was determined. Then the performance of K was predicted by the hybrid DNN-MBGD model and compared with the ANN (artificial neural network), SVM (support vector machine) and GMDH (group method of data handling). From the performance comparison and error analysis hybrid DNN-MBGD model result gives better performance than other models.