An interactive flood forecasting tool with ensemble-based machine learning model: A Bangladesh Perspective

Abstract

Flood is a major natural disaster that leaves a trail of destruction in its wake, destroying homes and infrastructure. In Bangladesh, frequent floods displace millions of people from their homes, disrupt their livelihoods, and cause loss of life every year. Conventionally, flood forecasting has relied on statistical analysis. Machine learning (ML) approaches have been increasingly explored for flood prediction and mapping in recent years. This paper aims to evaluate the effectiveness of several machine learning (ML) algorithms for detecting floods in Bangladesh. To achieve this, a rainfall dataset has been collected from multiple stations across Bangladesh at various periods. Besides, 11 ML algorithms have been developed to predict floods. Three metrics were used to evaluate the ML algorithms-accuracy, precision and F1 score. As a further contribution, the results obtained from the study are represented via an interactive tool to facilitate further research and analysis.