Detecting Flood in Real-Time Weather Data Using Anomaly Detection Algorithm

This paper presents flood detection using anomaly detection algorithms in real-time weather data. Nine different attributes are used to detect suspicious weather that can cause floods. The weather dataset was acquired from Kelantan, Malaysia. In this paper, the collected dataset will be sent from the chosen Internet of Things (IoT) pipeline to three different anomaly detection algorithms. These algorithms are multi-threaded autonomous anomaly detection (MAAD), robust random cut forest (RRCF), and outlier detection in feature-evolving data streams (xStream). Our evaluation demonstrates that the MAAD algorithm has a 68.5% precision score and a 0.01 false alarm rate, which is better than the RRCF and xStream for detecting floods.