Machine Learning Model Design for IoT-Based Flooding Forecast

Abstract

Flooding risk is a threat to sea-level residential areas in southern Sweden. An Internet of things (IoT) project has been deployed to monitor weather and water pipe conditions in Kristianstad, Sweden. The IoT data however only monitors the current condition and does not tell the future threat. Machine learning models using deep learning neural networks have been developed to predict future threats based on IoT data and weather forecast. This paper presents multiple model architectures and their performances. All the models are explainable. Finally, a conclusion is made by selecting the best-functioning model in the context of flooding risk prediction in Kristianstad.