PREDICTION OF DIABETES MELLITUS INSURANCE CLAIM MODELS USING MACHINE LEARNING METHODS

Abstract

Diabetes mellitus is an increase in blood sugar levels accompanied by impaired metabolism of carbohydrates, lipids, and proteins as a result of insufficient insulin function. In 2021 the number of deaths due to diabetes mellitus in Indonesia reached 236,711 people, this is ranked sixth in the world and ranked first in Southeast Asia. Also in Indonesia, this disease increased by 8.5% in 2014 in people over 18 years of age. Many factors influence this disease, including age, gender, as well as the doctor's diagnosis of congenital diseases. The increasing number of cases of death from diabetes mellitus every year causes insurance companies to anticipate the situation, including calculating appropriate claim reserves. This paper aims to calculate the prediction of claims that can be generated using the variable limits of age, gender, and doctor's diagnosis of other congenital diseases by doing classification which carried out using the K-Modes clustering and the Heuristic Method. After classifying the data, we  proceed with calculating claim predictions using Random Forest, Naïve Bayes, and Support Vector Machine algorithms. The results of this study indicate that the best model predictions are obtained using the Naive Bayes algorithm, while the best classification group uses the Heuristic model. This research will obtain the best accuracy if balanced with a large amount of data and more diverse variables. The results of this study are expected to be a guideline for insurance companies in determining the estimated amount of claims that may occur.