Water Level Prediction Based on Internet of Things and Machine Learning

Abstract

This study develops an Internet of Things and machine learning based water-level prediction system for Jalan Embun Suryana, Samarinda. The device uses an ESP32, a JSN-SR04T ultrasonic sensor to measure water-surface distance, and a tipping-bucket ombrometer for rainfall. Data are synchronized to Firebase and transformed into a simple accumulated-rainfall feature to train a simple linear regression model. The research stages comprised Requirements Analysis, System Design, Sensor Data Collection, Model Training, Model Testing, Device Testing, and Integration. An Android application provides real-time visualization and early-warning notifications. Experiments indicate good sensor accuracy (JSN-SR04T; tipping-bucket 0.7 mm/tip) and model performance (MSE = 0.562 cm², RMSE = 0.749 cm, R2 = 0.992). In sum, data acquisition, prediction, and notifications operated as designed, and the simple linear regression produced water-level forecasts. At the study site, the association between increasing rainfall and rising water level was corroborated by repeated manual field measurements.