An Integration of Least Squares Support Vector Machines and Firefly Optimization Algorithm for Flood Susceptible Modeling Using GIS

Abstract

The main aim of this research is to propose and evaluate a new hybrid intelligent approach (namely LSSVM-FA) based on Least Squared Support Vector Machines (LSSVM) and Firefly algorithm (FA) for flood sus- ceptible modeling with a case study at a typical flood region in Central Vietnam. LSSVM and FA are current state-of-the art machine learning techniques that have rarely been explored for flood study. For this aim, a geospatial database of flood for the study area was constructed that consists of 76 historical flooded locations and 10 influencing factors. Using the database, the flood model was established using LSSVM, and then, the model was optimized where the best model’s parameters were determined using FA. The goodness-of-fit and the prediction capability of the proposed model were evaluated using Receiver Operating Characteristic (ROC) curve and area under the ROC curve (AUC). The results showed that the proposed model performs well with the training data (AUC = 0.961) and the validation data (AUC = 0.934). Since the proposed model is better than benchmarks i.e. Neuron-fuzzy, support vector machines, and random forest, it could be concluded that the proposed model is a promising tool that should be used for flood modeling. The result from this research is useful for land-use planning and management at flood-prone areas.