OPTIMIZING MOTOR VEHICLE INSURANCE PURE PREMIUM DETERMINATION USING LOCALLY COMPENSATED RIDGE - MULTIVARIATE GEOGRAPHICALLY WEIGHTED REGRESSION MODEL Lisa S. Hasiru, Vicko R. Widodo, Kurnia Novita Sari, Natalie Calosa
Institute of Technology Bandung
Abstract
This research focuses on motor vehicle insurance, which provides protection against losses sustained by motor vehicles. Premiums for this type of insurance cannot be generally set for all policies but must be tailored to the individual risk profiles of each policyholder. Geographic factors play a crucial role in determining these risks. Utilizing claim data from a general insurance company in Indonesia, this research explores various geographic risk factors, including the number of motor vehicles such as motorcycles, cars, trucks, and buses, as well as road conditions and population density. These factors are analyzed in relation to claim frequency and severity. The predictive results from this analysis are used to calculate the pure premium for each policy. To model the relationship between risk factors and response variables, this research employs the Locally Compensated Ridge - Multivariate Geographically Weighted Regression (LCR-MGWR) model. This model extends the Geographically Weighted Regression (GWR) model by addressing multicollinearity issues in the independent variables within spatial data. The weights for parameters in this model are determined using the Gaussian kernel function for the claim frequency response variable and the Bisquare kernel function for the claim severity response variable. This research aims to provide deeper insights into geographical risk factors in motor vehicle insurance premium determination and offer a more accurate and fair approach to premium setting.
Keywords: Claim Frequency- Claim Severity- Locally Compensated Ridge - Multivariate Geographically Weighted Regression (LCR-MGWR)- Motor Vehicle Insurance- Pure Premium.
