MLP-Based Research in Flood Probability Prediction

Abstract

Flood is a kind of natural disaster with great harm. Probability prediction can reduce economic loss effectively. This paper optimizes the prediction of flood probability based on MLP model to reduce the loss. Based on the historical flood data set, a comprehensive correlation analysis using Pearson's correlation coefficient was conducted to find out the influence of each potential factor on the probability of flood occurrence. Based on this, the potential causes of flooding by these factors are examined in detail by taking into account various theoretical and practical aspects, with the aim of revealing the underlying mechanisms and patterns. Two different methods, linear regression and MLP model, are used to establish the corresponding flood probability prediction models. The classical linear regression method is firstly used to predict the probability of flood occurrence, and after rigorous calculation and verification, the obtained prediction accuracy is 99.9250%. In order to further optimize this prediction model to achieve higher accuracy and reliability, MLP is skillfully applied based on five key indicators. Through a series of calculations, the prediction accuracy was successfully increased to 99.9254%. This shows that the improvement is effective and provides a more accurate and efficient method for predicting the probability of flooding. This result not only verifies the rationality and effectiveness of the prediction model, but also provides a solid foundation and strong support for further research and application.