FLOOD PREDICTION USING MACHINE LEARNING

This study evaluates state-of-the-art machine learning (ML) models for flood prediction, focusing onrobustness, accuracy, effectiveness, and speed to determine the most suitable approaches. Aperformance comparison provides insights into various ML techniques for short- and long-termflood forecasting.Key strategies for improving flood prediction include hybridization, data decomposition, algorithmensemble, and model optimization. This research introduces hybrid models combining the Index ofEntropy (IOE) with Decision Tree (DT), Support Vector Machine (SVM), and Random Forest(RF) for Flood Susceptibility Modelling (FSM). Feature selection and multicollinearity analysisassessed the predictive capability of influencing factors. IOE was used to analyse flood-inducingfactors and assign weights for ML model training. Performance evaluation using the Area UnderCurve (AUC) showed that the DTIOE hybrid model achieved the highest accuracy (87.1%), whileDT alone had the lowest (77.0%). Hybrid models outperformed standalone models, andsusceptibility maps revealed that 21% of the study area is highly flood-prone, with human activitiessignificantly contributing to risk.