Published Paper Details:

Water is a vital resource for the survival of humans, animals, and plants. However, its quality is not always suitable for drinking or other uses due to factors like industrialization, pollution, and natural occurrences. The World Health Organization has guidelines to determine the threshold levels of various parameters in water samples. The Water Quality Index (WQI) and Irrigation WQI (IWQI) are used to assess overall water quality. Collecting and measuring water samples can be challenging, but a proposed network architecture uses real-time data collection and machine learning tools to automatically determine the suitability of water samples for drinking and irrigation. The network is based on LoRa and considers the land topology. Simulations indicate a partial mesh network topology is most suitable. Since there is a lack of large datasets on drinking and irrigation water, new datasets were developed for training machine learning models. Three models (Random Forest, Logistic Regression, and Support Vector Machine) were evaluated, with Logistic Regression performing best for drinking water and Support Vector Machine for irrigation water. Recursive feature elimination was used to identify the water parameters with the greatest influence on classification accuracies.