Predictive Model for Health Insurance Policy Renewal of Bank Customers

Currently, the insurance business is highly competitive. Retaining existing customers

has become a more effective strategy than acquiring new customers because it greatly reduces

the cost of marketing expenses. This research aims to create a model to predict a renewal of

health insurance policies of a bank's customers by using data from 2021 to September 2024,

totaling 697 policies. The factors studied include gender and age range of customers,

premiums, premium growth rate, insured sum, policy plans, use of hospital services in the

network, number of claims, and total claim amount. The researchers adjust the data balance

using the SMOTE technique before fitting prediction models using the elastic net logistic

regression and decision tree methods. The feature importance is used to determine the factor

affecting the decision of policy renewal. The research results reveal that the decision tree model

provides higher prediction efficiency, with an accuracy value of 0.9952, an F1-Score of 0.9818,

and an AUC of 0.9966. The main factors affecting the decision of policy renewal are the

insured amount at 55.09%, the Plan C policy at 10.41% and the original premium at 8.02%.

This shows that the level of coverage and the insurance product are the main variables

influencing the customer's decision. The results of this research can be applied for planning

marketing strategies, identifying customer groups at risk of not renewing the policy, and

adjusting customer retention strategies to suit each individual.