A flood risk assessment model based on Random Forest and its application

According to disaster system theory, 10 indexes are selected with an overall consideration of disaster-inducing factor, the hazard inducing environment and the hazard bearing body. An assessment model of flood risk is constructed based on an intelligent algorithm called Random Forest (RF). Taking a grid of 100 m×100 m as the computing element, the flood risk in Dongjiang River basin is evaluated by the model with the help of GIS technique. The evaluation results show that: the RF model has fewer parameters and dispenses with consideration of index weight and grading standard, making the implementation simpler. It can evaluate the index importance, which maks it convenient to analyze the contribution of each index to flood risk. It achieves higher accuracy, better classification result and stronger data mining ability than the SVM model. It is easy to integrate the GIS technique so that it can conveniently analyze the spatial pattern and inherent law of the risk. Therefore, the assessment model based on RF achieves good results and it can be a new way to flood risk assessment.