Modeling Fire Insurance Claim Frequency Using Negative Binomial Regression

Fire insurance plays an important role in providing financial protection against losses caused by fire risks. To support risk management and accurate premium determination, a model capable of predicting claim frequency based on relevant factors such as the Total Sum Insured (TSI) is required. The data used in this study consist of statistical fire insurance data covering the number of policies and claim frequencies in three provinces: West Java, Central Java, and East Java. The analysis was conducted using Poisson regression and Negative Binomial regression to model and predict claim frequency based on TSI. Initial estimation using the Poisson model indicated the presence of overdispersion, suggesting that this model is less suitable for the data. Therefore, the Negative Binomial regression model was applied, as it can better handle excessive variance. This model produced a lower AIC value compared to the Poisson model and showed that TSI has a significant effect on claim frequency. Thus, the Negative Binomial regression model is considered more accurate for predicting fire insurance claim frequency based on TSI.