COMPARISON OF GENERALIZED LINEAR MODEL AND GENERALIZED ADDITIVE MODEL IN DETERMINING HEALTH INSURANCE PREMIUM

• Main Author: ANJANI, WINIARDHITA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/65390
• Institution: Institut Teknologi Bandung

Health is a very important aspect, hence it is necessary to seek treatment for health problem both personally and in a hospital. Financial losses from that treatment can be transferred to health insurance. To cover all possible losses, the health insurance company must set a premium, which is the rate from the insurance contract that must be paid by the policyholder. Since each individual has a different risk of health problems, premiums cannot be set the same for all policyholders. The premium must be calculated by taking into account the risk factors of each policyholder. The impact of risk factors which are the predictor variables, namely age, gender, and type of insurance product, on the claim frequency and severity response variables will be modelled using data on health insurance claims of a company's employees. The expected frequency and severity of claims will result in a pure premium for each individual. In determining pure premium for health insurance, we will compare the ability of the Generalized Linear Model (GLM) and Generalized Additive Model (GAM), an extension of the GLM that allows to model the non-linear effect of the predictor variables on the response variable. In this Final Project, 8.745 policyholders with a total of 1.110 claims from 2019 health insurance data are modelled. The obtained GLM and GAM models can cover all policyholder claim losses. The GAM model outperforms the GLM model in determining pure health insurance premiums based on MAE and MDAE's evaluation of the claim severity and frequency prediction results.