Machine Learning Approach to IoT- Based Water Quality Monitoring

Abstract

Water resources are inevitable for human survival but untreated wastewater harms the environment. Thus, ongoing monitoring of water quality is necessary to identify pollution sources and prevent further damage. For such monitoring, an IoT water quality monitoring system was developed using Arduino technology to collect and transmit data to MQTT Brokers and store it in a database. The data is presented on a monitoring webpage. Three machine learning methods (Random Forest, ANN, and LightGBM) were used for backend analysis and prediction. LightGBM was found to have the highest prediction accuracy for NH3, pH, ORP, and temperature. The research contributes to reducing the need for frequent and costly data collection by using an IoT system for real-time monitoring and employing machine learning predictions to compensate for missing data. This approach provides a more efficient and effective method for analyzing and predicting water quality.