PEMODELAN PREMI ASURANSI BERDASARKAN DATA SEVERITY MENGGUNAKAN MODEL LOGNORMAL

Abstract

The insurance industry in Indonesia requires reliable quantitative approaches to accurately determine premium rates and manage claim risks effectively. This study aims to model pure insurance premiums based on claim severity data using the lognormal regression approach. The data used consist of historical individual claim amounts (severity) obtained from a general insurance company in Indonesia, covering the period from 2009 to 2015. Initial data exploration revealed that the distribution of claim values is positively skewed, indicating the suitability of lognormal modeling. Three models were evaluated: Generalized Linear Model (GLM) with Gamma distribution, GLM with Inverse Gaussian distribution, and linear regression with lognormal transformation. Model selection was based on the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). The results show that the lognormal model had the lowest AIC and BIC values, indicating superior performance compared to the other models. The selected model was then used to forecast pure premiums for the next 12 months, followed by the calculation of commercial premiums with a 30% loading factor. The prediction results show a consistent and proportional upward trend in premiums, demonstrating the model’s effectiveness in capturing historical claim patterns and supporting data-driven premium setting.