TIME SERIES ANALYSIS FOR CLAIM EVENTS IN PROPERTY INSURANCE THROUGH ZERO INFLATED POISSON AUTOREGRESSIVE APPROACH

• Main Author: Hudan Satrio, Rayhan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/69343
• Institution: Institut Teknologi Bandung

In property insurance, seeing how much the frequency of claims at a location occur over time is something interesting to research. Sometimes, filing a property insurance claim does not always happen on a regular basis every time. At certain intervals, there may be regular claims every day, but for other time intervals, there may be no claims at all. In this study, an analysis of time series data is carried out in the form of claims whose frequency is rare so that there are many zero values in the data. Generally, discrete data on the claim frequency are modeled by the Poisson model. However, if the modeling is carried out at the frequency of claims that rarely occur, the variance that will be generated from the data will be greater than the average, or commonly called overdispersion. As a result, it will not be suitable if it is modeled with the Poisson model. Thus, one alternative that can be used is to model the data with a Zero Inflated Poisson (ZIP) distribution. In building the ZIP Autoregressive model, whose claim frequency depends on the frequency of claims in the previous time, the Generalized Linear Model (GLM) method will be used. The application of the ZIP Autoregressive model is used for property insurance claim frequency data in DKI Jakarta province.