Forecasting of Flood in the Non-Tidal River of Northern Regions of Bangladesh Using Machine Learning-Based Approach

Abstract

Floods are among the most devastating natural disasters, causing extensive damage to property and posing a threat to human lives. However, significant progress has been made in mitigating their impact through the development of effective early warning systems. Over the past two decades, advances in machine learning (ML) technology have played a crucial role in enhancing the predictive capabilities of these systems. A recent study focused on predicting floods in non-tidal rivers by proposing various machine-learning models. The research findings indicate that the Random Forest algorithm emerges as the most effective, offering an accuracy of 87% with high precision, recall, and F1 scores, using an 80:20 training and testing data ratio. These findings provide valuable insights for hydrologists and make a significant contribution to flood forecasting and mitigation efforts. The study has significant implications for flood understanding and management, offering a better understanding of machine learning model performance in predicting floods in non-tidal rivers. This research provides a solid foundation for the development of more efficient early warning systems. The information gleaned from this study can be utilized by hydrologists, climate scientists, and other related practitioners to develop more accurate and reliable forecasting strategies in the face of flood threats. Thus, this research is not only a valuable scientific contribution but also a practical tool for future flood disaster risk mitigation efforts.