Deep learning approach for prediction and classification of potable water

Abstract

Potable water, commonly known as drinking water, refers to water that is safe to drink and does not endanger human health. It must adhere to strict quality standards set by health organizations, be devoid of dangerous pollutants and chemicals, and meet certain requirements for safety. The health of the public and the ecosystem are directly affected by water quality. Various pollutants have posed dangers to water quality in recent years. A more efficient and affordable approach is required due to the grave effects of low water quality. In this proposed research work, deep learning algorithms are developed to predict the water quality index (WQI) and water quality classifications (WQC), which are vital parameters that can be utilized to know the status of the water. To predict the WQI, a deep learning algorithm called long short-term memory (LSTM) is used. Further, WQC is performed using a deep learning algorithm called a convolutional neural network (CNN). The proposed system considers seven water quality parameters, namely, dissolved oxygen (DO), pH, conductivity, biological oxygen demand (BOD), nitrate, fecal coliform, and total coliform. The experimental results showed that the LSTM can predict water quality with superior robustness and predict WQI with the highest accuracy of 97%. Similarly, the CNN model classifies the WQC as potable or impotable with superior accuracy and a reduced error rate of 0.02. Graphical abstract: [Figure not available: see fulltext.]