Clssification model for demand of medical insurance using machine learning models

Abstract

This study aims to classify the demand for medical insurance in Uganda using machine learning models, specifically Naive Bayes and Random Forest. Data was collected from a brokerage firm in Kampala, focusing on demographic factors such as age, gender, marital status, occupation level, currency description, and basic premium. The Random Forest model significantly outperformed the Naive Bayes model, achieving 95.61% accuracy on the test set and demonstrating superior reliability in handling complex, imbalanced datasets. Key predictive factors identified were basic premium, age, and occupation level. The study successfully addressed several research questions, including the influence of age, gender, and marital status on insurance demand, as well as comparing the performance of Naive Bayes and Random Forest models. However, it did not fully explore the impact of currency description or the implications of different occupation levels. Recommendations include investing in technology and data infrastructure, enhancing public awareness, fostering collaboration between stakeholders, developing supportive policy frameworks, and conducting further research on unexamined factors. The findings suggest that adopting the Random Forest model can significantly improve the efficiency, accessibility, and affordability of medical insurance in Uganda, providing a model for other regions with similar challenges.