Mobile Based Flood Disaster Detection Using K-Nearest Neighbors Algorithm Yaqutina Marjani Santosa (a*), Nur Budi Nugraha(a), Alifia Puspaningrum(a)
(a)Department of Informatics Engineering, Politeknik Negeri Indramayu, West Java, 45252, Indonesia
Abstract
Flood disaster is one of the most frequent disasters in the world. One of the main challenges in flood disaster detection is the limitation in accurately predicting when, where, and how severe a flood will occur. In this study, a mobile-based flood status detection system is proposed that utilizes the k-Nearest Neighbors (kNN) algorithm. The proposed approach will integrate various variables such as real-time rainfall data, topographic information, land use data, and historical flood records to produce an accurate and responsive prediction model. In addition, this study will also explore ensemble techniques by combining kNN and other machine learning algorithms to improve prediction performance. The trial demonstrated the effectiveness of the system, achieving 93.5% accuracy on the test data. Comparative analysis showed that kNN is on par with other machine learning algorithms in detecting flood status, with advantages in terms of interpretability and ease of implementation on mobile devices. The performance of the mobile application is very promising, with an average prediction time of 0.5 seconds, indicating its suitability for direct use. The results of this study are expected to make a significant contribution to the development of a more effective, adaptive, and easy to implement flood early warning system in various geographic contexts.
