Bayesian Predictive Analysis of Natural Disaster Losses

Different types of natural events hit the United States every year. The data hazards from 1900 to 2016 in US shows that there is an increasing trend annul disaster losses after 1980. Climate change recognized as one factors causing this trend, and predictive analysis becomes important loss prediction risk prevention continues. In paper, we convert year dollars using yearly average Consumers Price Index (CPI), conduct several tests verify CPI adjusted amounts individual disasters are independent identically distributed. Based on these test results, use various model selection quantities find best for severity among three composite distributions, namely Exponential-Pareto, Inverse Gamma-Pareto, Lognormal-Pareto. These distributions piecewise small with high frequency large low frequency. Remarkably, make first attempt derive analytical Bayesian estimate Lognormal-Pareto distribution based selected priors, show outperforms other two modeling losses. Important measures thereafter derived discussed.