Flood susceptibility mapping using support vector regression and <scp>hyper‐parameter</scp> optimization

Floods are both complex and destructive, in most parts of the world cause injury, death, loss agricultural land, social disruption. Flood susceptibility (FS) maps used by land-use managers land owners to identify areas that at risk from flooding plan accordingly. This study uses machine learning ensembles produce objective reliable FS for Haraz watershed northern Iran. Specifically, we test ability support vector regression (SVR), together with linear kernel (LK), base classifier (BC), hyper-parameter optimization (HPO), flood-prone this watershed. We prepared a map 201 past floods predict future floods. Of flood events, 151 (75%) were modeling 50 (25%) validation. Based on relevant literature our field survey area, 10 effective factors selected zoning. The results show three important predicting flood-sensitive areas, specifically order importance, slope, distance river river. Additionally, SVR-HPO model, area under curve values 0.986 0.951 training testing phases, outperformed other two tested models.