Development of a novel hybrid multi-boosting neural network model for spatial prediction of urban flood

In this study, a new hybridized machine learning algorithm for urban flood susceptibility mapping, named MultiB-MLPNN, was developed using multi-boosting technique and MLPNN. The model tested in Amol City, Iran, data-scarce city an ungauged area which is prone to severe inundation events currently lacks prevention infrastructure. Performance of the compared with that standalone MLPNN model, random forest boosted regression trees. Area under curve, efficiency, true skill statistic, Matthews correlation coefficient, misclassification rate, sensitivity specificity were used evaluate performance. validation, MultiB-MLPNN showed best predictive thus useful generating realistic maps areas. can be develop risk-reduction measures protect areas from devastating floods, particularly where available data are insufficient support physically based hydrological or hydraulic models.