M3MLF: Mapping, Measuring and Monitoring Using the Machine Learning Framework for Dynamic Water Source Classification

Water is the most valuable natural resource for all living organisms. Ever-increasing population, climatic variability, and anthropogenic activities are exacerbating the world’s water crisis. However, the community members and local administration lack the knowledge of dynamically varying water contamination levels and public health impacts due to this change. The proposed work is aimed to design an IoT system for real-time monitoring of water quality in geographically separated heterogeneous water sources. Support vector-based classification is implemented to understand the dynamic water quality variability of spatially distributed heterogeneous water resources from Thiruvananthapuram, Kerala, India. The sources were mapped into any of the three classes such as Good, Poor, and Unsuitable, and SVM based classification is performed. The results show very good precision and recall in predicting the water quality index class for the test data. To enhance community empowerment, a public portal is developed for easy visualization and information dissemination to multiple stakeholders. The portal provides spatial trends, temporal trends for an area or a particular location, area-wise prevalence of selected parameters, determination of water quality index, classification of specific sources based on water quality index, and proposing the required treatments for the sources based on its water quality index.