Hydro Sentry: IoT-Based River Water Level Monitoring and Forecasting Using Dual Layer Neural Networks

The Philippines is an archipelago nation particularly vulnerable to natural disasters. According to the Flood Hazard Maps provided by the Philippine Atmospheric, Geophysical, and Astronomical Services Administration (PAGASA), several river basins are highly vulnerable to flooding. The increasing frequency and severity of flooding events due to climate change and rapid urbanization have necessitated the development of efficient and reliable river monitoring and forecasting systems. Accurate and timely information regarding river water levels is crucial for flood prevention, disaster management, and safeguarding communities living in vulnerable areas. Traditional methods of river monitoring, which often rely on manual observations and simplistic prediction models, are limited in their ability to provide real-time data and accurate forecasts. To address this pressing issue, this paper proposes the development of an IoT and web-based water level monitoring and forecasting system. The forecasting model utilizes a 128-neuron ReLU-activated dense layer for feature extraction followed by a single-neuron linear layer for final prediction. The Adam optimization algorithm ensures efficient model training. The Mean Square Error (MSE) quantifies the average squared differences between predicted and actual values, offering an insight into the overall deviation of the model's predictions. This metric is particularly sensitive to larger errors due to the squaring of discrepancies before averaging. The dual-layer forecasting model demonstrated an MSE of 7.45 during the training phase and 12.56 during the testing phase, indicating that the prediction errors are relatively moderate. s Overall, system aims to empower these flood-prone communities by providing them with real-time data and proactive flood predictions.