Scholarlink

The potential for unresolved flooding in the Kali Bekasi area, West Java, Indonesia will pose a risk to technical activities in the Kali Bekasi area project which is being carried out by the ADHI as the contractor. So, a system is needed that can detect floods to reduce obstacles to project implementation. This research develops a flood prediction system that uses an artificial neural network approach on the backend. The Raspberry Pi connects to an Internet of Things (IoT) device to measure the distance to the water surface. Ultrasonic sensors and webcam modules take photos of river conditions for data processing. Extensive testing of the flood prediction system demonstrated reliability and accuracy. The HC-SR04 ultrasonic sensor achieved high precision (97.95%–99.32%) in measuring water levels, while the webcam effectively captured and transmitted river images to the database. The web application performed seamlessly across all features during testing. Machine learning experiments using deep learning algorithms showed excellent predictive performance, with a minimal Mean Squared Error (MSE) of 0.038%, highlighting the system’s effectiveness for early flood detection and management.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Flood Disaster Early Prediction System Using Machine Learning Based on Internet of Things (IoT) on the ADHI Construction Project

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ratih Dwi Anggraeni,
  • Fahzrin Perwitasari,
  • Listyo Edi Prabowo,
  • Raymond Luthfi,
  • Sutarto

摘要

The potential for unresolved flooding in the Kali Bekasi area, West Java, Indonesia will pose a risk to technical activities in the Kali Bekasi area project which is being carried out by the ADHI as the contractor. So, a system is needed that can detect floods to reduce obstacles to project implementation. This research develops a flood prediction system that uses an artificial neural network approach on the backend. The Raspberry Pi connects to an Internet of Things (IoT) device to measure the distance to the water surface. Ultrasonic sensors and webcam modules take photos of river conditions for data processing. Extensive testing of the flood prediction system demonstrated reliability and accuracy. The HC-SR04 ultrasonic sensor achieved high precision (97.95%–99.32%) in measuring water levels, while the webcam effectively captured and transmitted river images to the database. The web application performed seamlessly across all features during testing. Machine learning experiments using deep learning algorithms showed excellent predictive performance, with a minimal Mean Squared Error (MSE) of 0.038%, highlighting the system’s effectiveness for early flood detection and management.