Flood Occurrence Probability Modeling and Evaluation Based on Spearman Analysis and MLP Algorithm

Abstract

Global climate change and human activities have intensified flood risk. Therefore, accurate prediction of flood probability and disaster loss reduction has become the focus of disaster prevention and mitigation research. At the same time, traditional statistical methods are difficult to meet the needs of flood prediction in complex environments. Given this, this study starts with many index factors that affect the probability of flood occurrence and combines the multi-layer perceptron model to predict the likelihood of flood occurrence. This model has a good effect in the field of prediction and can make up for the shortcomings of traditional statistical methods. Specifically, this paper first conducts Spearman correlation analysis for different indicators, aiming to select 11 indicators with high correlation. The multi-layer perceptron model was further used to modify the model parameters and data set division. Finally, the flood probability prediction model was obtained with a 60% training set and 40% verification set, the number of hidden layer neurons was 64 and 32 respectively, and the root mean square error RMSE was 0.038242. The model was considered reliable. Based on this, it is applied to the event prediction of unknown flood occurrence probability, and the results of the event prediction are 0.534, 0.470, 0.452, etc., and most of them fall between 0.45 and 0.55. At the same time, this study deeply explored the distribution of flood probability, combined with the Q-Q diagram and probability distribution, and found that the data set conforms to the normal distribution, which is consistent with the real world.