Classification of Tropical Cyclone Risks in the Philippines using Random Forest

Abstract

The Philippines experienced an average of twenty (20) tropical cyclones every year. With the aim to help the government in mitigation of the potential impact of the tropical cyclones in the country, this research explored the classification of risk brought about by the said natural calamity. Due to the excellent performance of Random Forest in various studies, this ensemble method was used in the risks classification. Data gathered from different government agencies were used as predictors or classifiers of the risk level of Tropical Cyclones. The research used the Exponential Regression for missing value imputation and converted the number of casualties, damaged houses and properties into five (5) risk levels using Quantile Method. The cleaned data were distributed into 80:20 ratios for training and testing sets respectively. The recorded optimal accuracy based on the experiment is approximately 93%, 75%, and 84% with average running time of 10.183s, 8.793s, and 8.245s for casualties, damage houses and damage properties respectively.