Machine learning approach for determining the water quality of freshwater lakes: A case study on selected lakes of Chennai, India

Abstract

AbstractMonitoring freshwater quality in urban like Chennai is indispensable as the chief drinking water source for the city originates from freshwater lakes. This study portrays an economical, user‐friendly, economical method to detect water quality. A machine learning‐based portable water quality check test kit was developed along with a user interface. To validate the efficiency of the portable kit, this study focussed on the ten major freshwater lakes of Chennai City. Basic quality parameters of water, such as the potential of hydrogen (pH), level of ammonia, nitrite, nitrate, and total phosphate (T.P.), were considered when designing this mobile phone‐compatible user interface application. The experiment used and tested water samples from ten different lakes with the commercially available freshwater master test kit. The images of color developed during the test were captured, and corresponding Red, Green, Blue (RGB) values were calculated. These values were used in Igray, Iavg, and Euclidean distance algorithms to predict the pollution level of the samples. Based on the models, the Euclidean model fitted well with the calibration data with a correlation value of more than 0.98 for all the parameters. Percent accuracy values of the methods used in this study were compared, and the Euclidean model showed more than 98% accuracy for all parameters. The result of this study is alarming that all the ten freshwater lakes considered in this study were hypereutrophic.