Medical Insurance Forecasting Using Neural Network
Ms. Melony Bharucha1, Ms. Jahanvi Mistry 2, Asst Prof. Ms. Manisha Vasava3
12Research Scholar, Department of Information Technology, Krishna School of Emerging Technology & Applied Research, KPGU University, Varenama, Vadodara, Gujarat, India
3Assistant Professor, Department of Information Technology, Krishna School of Emerging Technology & Applied Research, KPGU University, Varnama, Vadodara, Gujarat, India
---------------------------------------------------------------------***---------------------------------------------------------------------
ABSTRACT
The accurate prediction of medical insurance costs plays a crucial role in effective healthcare planning, insurance Premium estimation, and risk assessment. Traditional statistical techniques often fall short in capturing the nonlinear relationship inherent in insurance data. In this research, we propose a machine learning-based approach for forecasting medical insurance charges by leveraging a Neural Network model and comparing its performance with a Random Forest algorithm. The dataset utilized includes demographic and lifestyle features such as age, sex, BMI, smoking status, number of children, and region. While the Random Forest model provides a solid baseline with good predictive accuracy and interpretability, our findings demonstrate that the Neural Network model significantly outperforms it in terms of predictive performance, achieving lower Mean Squared Error(MSE) and higher R2 scores. The neural architecture is capable of learning complex feature interactions, making it especially suitable for healthcare-related cost estimation tasks. our research extends the analytical framework by integrating deep learning for improved prediction precision. The study concludes with an evaluation of the practical implications of neural network based insurance cost forecasting for actuaries, insurers, and policyholders, and highlights avenues for future research incorporating larger datasets and advanced deep learning architectures.
Key Words: Neural Networks, Random Forest regression