Floodcast 2.0: real-time flood level monitoring and forecasting network through internet of things (iot) and machine vision system with web-based interface/ Kian N. Bataclan, January R. Beljot, Raymond Karl L. Lapating, Camila Ann P. Llemos, and Jeremiah John C. Uy.--

Bachelor's thesis

College Of Engineering.-- Bachelor of science in electronics engineering: Technological University of the Philippines,
2025.

Includes bibliographic references and index.

The Philippines is highly prone to natural disasters because of its location, making
flooding a frequent problem in areas like Bulacan. This study developed a real-time flood
level monitoring, prediction, and alert system to help inform Barangay Francisco Homes -
Yakal and other nearby residents in Bulacan about potential flooding. The system
integrates various hydrological sensors, including a submersible pressure level sensor, an
ultrasonic sensor, a rain gauge, and a water flow sensor to collect hydrological data. This
data is transmitted in real time using Internet of Things (IoT) technology. For predicting
floods, the Long Short-Term Memory (LSTM) model gave the best results with a Mean
Absolute Percentage Error (MAPE) of 0.64%, performing better than XGBoost, Random
Forest, and ARIMA models. Water level detection using a YOLOv7 model trained with
Roboflow 3.0 reached 99.5% mean Average Precision at 0.5 (mAP@50), achieving 100%
precision and recall even under different weather conditions. The system measured water
levels with 98% to 99% accuracy during the day and 97% accuracy at night using an
infrared IP camera. Sensor calibration resulted in an average error of less than 2%, which
proved its reliability. A web-based interface was developed to display real-time flood data,
provide early warnings, and generate emergency information to residents and local
authorities.