Analysis and Prediction of Flood Disasters Based on BP Neural Network and Convolutional Neural Network

This paper presents an analysis and prediction of flood disasters based on the BP neural network and convolutional neural network. Initially, employing the Spearman correlation test, all indicators for flood prediction were selected. Subsequently, a prediction model based on the BP neural network was established, and flood data were randomly divided into training, validation, and test sets for training. The model's convergence curve and a coefficient of determination of 0.883 demonstrated its strong performance. Additionally, employing the reverse BP neural network, five indicators with the highest importance scores were identified. The convolutional neural network was then used to predict the probability of flood occurrences, resulting in a coefficient of determination of 0.837, indicating the model's ability to maintain high accuracy even with fewer features. Furthermore, flood prediction was conducted based on all features and the convolutional neural network, achieving a coefficient of determination of 0.909 on the validation set.