Flood Susceptibility Prediction Using Machine Learning: A Case Study of the Ghaghra Basin

The area of the Ghaghra Basin faces a frequent flood event, and with the increasing frequency and intensity of flood events, it has highly impacted human lives, infrastructure, and agriculture. Hence, there is a pressing need for accurate and timely flood prediction models. Traditional hydrological models are useful but often inaccurate for effective disaster management. The new technologies, like machine learning, are a prominent solution for complex problems like floods. Therefore, this study conducts a comprehensive analysis of advanced machine learning techniques: logistic regression, support vector machines (SVM), random forest (RF)and XGBoost, in predicting floods in the Ghaghra Basin by using GIS, DEM, remote sensing data and seven flood-influencing factors. The performance of the models is tested by using various performance metrics such as accuracy, recall, F1 and precision. Further, the models were also validated using historical flood data by using the AUC-ROC curve. The results reveal that XGBoost outperforms the other models in prediction accuracy. Logistic regression and SVM also perform well, giving them viable options depending on the application scenario. This study examines the benefits and weaknesses of each model and recommends their use in flood-prone areas like the Ghaghra Basin.