A1302
Title: Grey and Markov Chain Grey modelling for insurance loss forecasting
Authors: Andreas Makrides - University of the Aegean (Cyprus) [presenting]
Alex Karagrigoriou - University of Piraeus (Greece)
Nikolaos Galanopoulos - University of the Aegean (Greece)
Abstract: Forecasting insurance loss data represents an important problem in actuarial and statistical modelling. An investigation of the predictive performance of the Grey Model GM(1,1) and the Markov Chain Grey Model MCGM(1,1) is conducted using state-level auto insurance incurred loss data from the United States. Forecast accuracy is evaluated using the Mean Absolute Percentage Error (MAPE). The empirical findings indicate that the MCGM(1,1) provides substantial improvements in predictive accuracy compared with the standard GM(1,1). The study also investigates the role of the background coefficient in the construction of the background sequence. The results suggest that optimal values of the background coefficient frequently occur near the boundaries of the interval [0,1], rather than the empirical Mean value commonly adopted in the literature.
