Predictive Modeling for Healthcare Insurance Costs Using Machine Learning

Abstract

Insurance policies help to reduce financial losses by covering various risks, including medical expenses. The task of grading academic assignments is typically cumbersome, unequal, and also charged with some human bias due to personal judgment especially when it is subjective, for instance, essays and short hand answers. The paper presents an AI grading system that automates both objective and subjective assignment grading based on state-of-the-art technology. The system includes Optical Character Recognition (OCR) for processing handwriting, NLP models for evaluating essays and textual responses, and machine learning algorithms for objective questions in multiple-choice and fill-in-the-blank formats. This system also delivers detailed feedback to improve learning outcomes. After significantly reducing grading time while remaining fair and accurate, this system presents a scalable and efficient solution for modernization in educational evaluation processes.

Introduction

Out-of-pocket payments dominate healthcare financing in many developing countries, creating barriers to universal health coverage due to inefficiencies, inequities, and high costs. Health insurance is essential for managing financial risks related to healthcare, but high premiums often leave many uninsured, delaying care and increasing mortality.

Accurately predicting individual healthcare expenses is critical for insurers, healthcare providers, and patients to optimize resource allocation, plan appropriately, and select suitable insurance plans. However, predicting costs is complex because medical events are often rare and vary across populations, necessitating fair premium models that consider individual factors.

This study uses demographic and behavioral data to build healthcare cost prediction models, comparing four machine learning techniques: Linear Regression, K-Nearest Neighbors (KNN), Support Vector Regression (SVR), and Random Forest (RF). Results show Random Forest performs best in accuracy and generalization, illustrating machine learning’s value in forecasting expenses, especially for high-cost patients, thereby supporting resource management and risk mitigation in insurance.

Literature Review:
Previous research highlights various machine learning models applied to insurance cost prediction using datasets with demographic and lifestyle attributes. XGBoost, Random Forest, and Gradient Boosting often perform well, though accuracy varies by model and data. Explainable AI techniques like SHAP help interpret model predictions, improving trust and decision-making. Studies also emphasize data preprocessing, feature selection, and model interpretability.

Related Work:
Machine learning models outperform traditional linear methods by capturing complex, nonlinear interactions in healthcare data, improving predictions in areas like disease progression and insurance fraud detection. This study uniquely compares multiple models to identify the best approach for insurance cost prediction.

Data and Methods:
The dataset includes features such as age, sex, BMI, number of children, smoking status, and region, with insurance charges as the target. Data preprocessing involves encoding categorical variables and scaling numeric features. Four models were trained and evaluated using 5-fold cross-validation, with performance measured by R², Mean Squared Error (MSE), and Mean Absolute Error (MAE). Random Forest showed superior predictive performance.

Conclusion

This study compared the performance of four machine learning models for predicting medical insurance charges. The Random Forest Regressor proved to be the most effective model, providing the highest R² score and the lowest Mean Squared Error. These results suggest that Random Forest is well-suited for cost prediction tasks involving non-linear relationships between features. Future research could focus on further enhancing model performance by incorporating more advanced techniques like XGBoost or Gradient Boosting, and feature engineering to derive new features from the existing dataset. Additionally, hyperparameter tuning for models like SVR and Random Forest could improve their predictive accuracy.

References

[1] Sazzad Hossen “Medical Insurance Cost Prediction Using Machine Learning“. October 2023 DOI:10.13140/RG.2.2.31456.25604 Thesis for: Medical Insurance Cost Prediction [2] Dr. S. M. Iqbal, Sayali D. Ghatol, Prerana V. Jadhav, Nikita D. Raspalle, “Health Insurance Cost Prediction Using Machine Learning” . [3] Kashish Bhatia, Shabeg Singh Gill, Navneet Kamboj, Manish Kumar,Rajesh Kumar Bhatia, “Health Insurance Cost Prediction using Machine Learning”. [4] Md Mohtaseem Billa, Dr. Tapsi Nagpal, “Medical Insurance Price Prediction Using Machine Learning“. [5] Ugochukwu Orji, Elochukwu Ukwandu, “Machine learning for an explainable cost prediction of medical insurance”. [6] S. Panda, B. Purkayastha, D. Das, M. Chakraborty and S. K. Biswas, \"Health Insurance Cost Prediction Using Regression Models”. [7] T. T, S. H. T, V. K. V and K. R, \"Medical Insurance Cost Analysis and Prediction using Machine Learning,\" 2023 International Conference on Innovative Data Communication Technologies and Application (ICIDCA), Uttarakhand, India, 2023, pp. 113-117, doi: 10.1109/ICIDCA56705.2023.10100057 [8] R. D, M. S. K and D. J, \"Health Insurance Cost Prediction using Machine Learning Algorithms,\" 2022 International Conference on Edge Computing and Applications (ICECAA), Tamilnadu, India, 2022, pp. 1381-1384, doi: 10.1109/ICECAA55415.2022.9936153. [9] A. Vinora, V. Surya, E. Lloyds, B. Kathir Pandian, R. N. Deborah and A. Gobinath, \"An Efficient Health Insurance Prediction System using Machine learning,\" 2023 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES), Chennai, India, 2023, pp. 1-5, doi: 10.1109/ICSES60034.2023.10465334. [10] Kaushik K, Bhardwaj A, Dwivedi AD, Singh R. Machine Learning-Based Regression Framework to Predict Health Insurance Premiums. Int J Environ Res Public Health. 2022 Jun 28;19(13):7898. doi: 10.3390/ijerph19137898. PMID: 35805557; PMCID: PMC9265373. [11] Sahu, Ajay and Sharma, Gopal and Kaushik, Janvi and Agarwal, Kajal and Singh, Devendra, Health Insurance Cost Prediction by Using Machine Learning (February 22, 2023). Proceedings of the International Conference on Innovative Computing & Communication (ICICC) 2022. [12] Mukund Kulkarni, Dhammadeep D. Meshram, Bhagyesh Patil, Rahul More, Mridul Sharma, Pravin Patange “Medical Insurance Cost Prediction using Machine Learning” [13] Uber Rides Prediction using Machine Learning Lokesh S Khedekar, Ajay S Chhajed, Ravishankar C Bhaganagre, Naina S Kokate, Swaraj Patil-2025 International Conference on Electronics and Renewable Systems (ICEARS) DOI:10.1109/ICEARS64219.2025.10940984 [14] Creative sustainability: Transforming household waste in India–A public survey on awareness and participation, L Khedekar, A Pandit, R Apte, B Adke, P Ambade, A Aher,Challenges in Information, Communication and Computing Technology DOI:10.1201/9781003559085-7 [15] Team portal website: Development & constructing bridges between teams,L. Khedekar, R. Dane, N. Dgama, S. Dangat, R. Dagade, V. Dahatonde, Challenges in Information, Communication and Computing Technology DOI:10.1201/9781003559085-30 [16] Innovating Healthcare: Developing a Comprehensive Patient Record Tracker System for Enhanced Medical Data Management and Patient Care,Lokesh Khedekar, Atharva Dhananjay Mohite, Arnav Meghan Kamat, Arpit Anil Topugol, Purva Dipak Atale, Pranay Suresh Asniyekar,SSRN 5086771 [17] AgriTech: Technology Driven E-Commerce Platform for Sustainable Agricultural Development ,Lokesh Khedekar, Radhika Dagade, Vaibhav Dahatonde, Rohit Dane, Sanskar Dangat, Prem Deore, Nevan Dgama, 2024 8th International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud)(I-SMAC)

Copyright

Copyright © 2025 Lokesh Khedekar, Bhawana Rakshit, Prutha Sawale, Siddhant Wasnik, Yashraj Shinde. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.