The Prediction Analysis of Zero Inflated Poisson Autoregression Model for the Number of Claims in General Insurance

The number of claims happen in a fixed interval time, is highly possible to be a rare event. It make the series data has many zeros frequency, and the variance data is much higher than its mean, which called as over dispersion. Commonly, to model the probability distribution of frequency data, the poisson distribution is favorable. However for over-dispersed model, it no longer appropriate. The Zero Inflated Poisson (ZIP) Autoregression be the strong candidate to solve it. This model offer prediction of upcoming count data through its probability distribution. Here, this prediction method is equipped with the analysis of cumulative distribution function behaviours which assumed to follow beta distribution. Through this approach, the at most upcoming count data can be predicted as a single number instead of its probabilty distribution. For case study, the number of general insurance happen in Jakarta City is used.