Classification and Hazards of Arsenic in Varanasi Region Using Machine Learning

Abstract

Groundwater plays a significant role in sustaining life in terrestrial and marine ecosystems. Arsenic contamination in aquifers poses a serious threat to the ecosystem due to its carcinogenic effect. Arsenic contamination in aquifers of the Varanasi region was noticed after water sampled from random sites of the Varanasi region of Uttar Pradesh, India. Under the Capacity Building of Urban Development (CBUD) scheme, Varanasi was chosen by the Ministry of Housing and Urban Poverty Alleviation (MoHUPA) and the Ministry of Urban Development (MoUD). In this study, various machine learning classifiers have been developed to classify water samples collected from the Varanasi region as safe or unsafe for consumption. The water with less than 10 µg/L As concentration is termed safe per World Health Organisation (WHO). Firstly the water samples parameters were ranked then the samples were trained and tested. Various parameters obtained from confusion matrices such as accuracy, precision, and recall are used to analyze the performance of different machine learning classifiers like Simple Logistic, MLP Classifier, and Random Forest. Among these models, Simple Logistic outperforms other classifier models. The Simple Logistic algorithm was considered the best model among the different classifiers. It has the highest accuracy of 79.03%, the highest precision of 77.00%, the highest recall of 79.00%, and a high ROC area of 69.40%. Thus, this model can be used for classification, and policymakers may devise plans to tackle the As poisoning in the Varanasi region.