Development of Combined Heavy Rain Damage Prediction Models with Machine Learning

Changhyun Choi,Hung Soo Kim,Jeonghwan Kim,Jungwook Kim

doi:10.3390/w11122516

Abstract

Adequate forecasting and preparation for heavy rain can minimize life and property damage. Some studies have been conducted on the heavy rain damage prediction model (HDPM), however, most of their models are limited to the linear regression model that simply explains the linear relation between rainfall data and damage. This study develops the combined heavy rain damage prediction model (CHDPM) where the residual prediction model (RPM) is added to the HDPM. The predictive performance of the CHDPM is analyzed to be 4–14% higher than that of HDPM. Through this, we confirmed that the predictive performance of the model is improved by combining the RPM of the machine learning models to complement the linearity of the HDPM. The results of this study can be used as basic data beneficial for natural disaster management.

Highlights

The intensity and frequency of extreme events has increased worldwide due to global warming [1,2,3,4]
To relieve the multicollinearity problem of the heavy rain damage prediction model (HDPM) developed from the linear regression model, we considered principle component analysis (PCA) that reduces the dimension of independent variables that have high correlation to an independent, small number of principal components
To evaluate the predictive performance of the HDPM and the combined heavy rain damage prediction model (CHDPM) that are modeled with the residual prediction model (RPM), this study used a test dataset that accounts for 30% of the total dataset and has not been used for model development

Summary

Introduction

The intensity and frequency of extreme events has increased worldwide due to global warming [1,2,3,4]. Studies that developed prediction models for natural disaster damages include linear regression models using independent variable like maximum wind speed, movement speed of typhoons, antecedent rainfall, total rainfall, and snowfall [9,10,11,12,13,14]. Some studies pursued to improve predictive performance by including social and economic factors as independent variables such as area, population, income level, number of houses, and financial independence [15,16,17,18,19,20]. Research trials were made using weather data as independent variables while grouping the social and economic characteristics of a study area to provide a damage prediction function [21,22]. Most of the past studies have used a single model, but recently studies are showing more interest in combining models with different characteristics to improve the predictive performance [33,34,35]

Objectives

Methods

Results

Conclusion