DOI: https://doi.org/10.21203/rs.3.rs-1166542/v1
The lack of a universal system for analysis, prediction, and storage of water quality and condition of rivers in Madhya Pradesh has led to uneven policy-making and poor management ultimately posing issues in health, irrigation and keep increasing pollution in rivers. This study is a part of developing a central system for river water quality assessment and prediction. The conventional method of water quality assessment is based on the calculation of the water quality index which can be very complex and time-consuming. This paper aims to develop a water quality prediction model with the help of an Artificial Neural Network (ANN) for predicting the water quality of the Narmada River using two machine learning algorithms Levenberg and Gradient Descent and the results were compared. This research uses the surface water historical data of years 2018, 2019 of the river Narmada with monthly time intervals. Data is obtained from the Central Pollution Control Board resource called Narmada Automatic Sampling Collection Stations System. For training the network 10 water quality parameters including, DO, BOD, Turbidity, pH, etc. After training the networks were accessed using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Coefficient of Correlation (R) out of which 2 best performing networks with 7 ( Training R = 0.80083, Testing R = 0.5767) and 19 (Training R = 0.6594, Testing R = 0.7424) Neurons in the hidden layer, were selected from Levenberg algorithm and, 5 (Training R = 0.7670, Testing R = 0.8123) & 17 (Training R = 0.8631, Testing R = 0.8981) Neurons in the hidden layer were selected from Gradient descent algorithm. This simplifies the calculation of WQI take care if any sampling station is out of service and data is not available for some reason. Further, the aim is to refine the prediction location-wise to be able to make a better decision when & where to implement the measures to reduce the pollution or the knowledge level of treatment required to make the water fit for use beforehand. This would be helpful in the treatment of water for use in Domestic or Irrigation Purposes.