DETERMINING VEHICLE INSURANCE PREMIUM USING LOCALLY COMPENSATED RIDGE-MULTIVARIATE GEOGRAPHICALLY WEIGHTED REGRESSION

• Main Author: Calosa, Natalie
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/82345
• Institution: Institut Teknologi Bandung

Motor vehicle insurance is a type of insurance that provides protection against losses experienced by motor vehicles. Premiums in motor vehicle insurance cannot be uniformly set for policies but must be adjusted according to the individual risks of each policyholder. One crucial factor to consider is the geographical aspect. By using claim data from a general insurance company in Indonesia, this research examines several risk factors related to the geographical aspect, such as the number of motorcycles, cars, trucks, buses, the percentage of national roads, provinces, unstable districts, and population density, against response variables such as claim frequency and severity. The predicted results of claim frequency and severity will then be utilized to generate pure premiums for each policy. To model the relationship between predictor variables and response variables, with a specific focus on the geographical or regional aspect, this research employs Locally Compensated Ridge – Multivariate Geographically Weighted Regression (LCR-MGWR). This model is an extension of Geographically Weighted Regression (GWR) with additional considerations for cases of multicollinearity in the independent variables of spatial data. The case study in this thesis uses policyholder data from the year 2016, with a total of 737,376 policies and 5,265 claims. The research findings indicate that the data used show spatial effects and multicollinearity, and the resulting pure premiums can cover the company's losses due to claims and provide a 47% profit to the company. Additionally, regional categories based on premiums have been determined, with provinces in the same category having similar premium amounts.