Prediction of Insurance Claim Payouts in Malaysia with State Space Modelling

Abstract

Insurance claim payouts represent a primary determinant in actuarial pricing models used to calculate policy premiums. However, high volatility in claims directly influences the accuracy of premium determination. Therefore, predicting insurance claim payouts is of critical importance to insurance companies for maintaining solvency. This paper examines the prediction of insurance claim payouts in Malaysia, specifically within the total permanent disability category, using a state-space model. The raw data undergoes pre-processing through cleaning and scaling to ensure consistency. A state-space model is then formulated to capture the underlying dynamics of claim payouts. Subsequently, a prediction-error-based cost function is introduced to estimate model parameters, with the output and state-transition matrices are updated via the gradient descent method. The performance of the model is assessed using a mean squared error (MSE). The results demonstrate that the state-space model closely tracks the trend of disability claim payouts, yielding a significantly lower MSE compared to traditional Simple Moving Average (SMA) and Exponential Smoothing (ES) models. In conclusion, the state-space model provides an efficient computational framework for claim prediction, proving to be a robust tool for actuarial forecasting in the Malaysian insurance sector.