Deep Learning-Based Medical Insurance Cost Prediction and Analysis

Abstract

The rapid advancement of technology is expanding its application possibilities across various human activities. In the healthcare sector, technologies such as the Internet of Things (IoT) support numerous applications, including disease analysis, smart monitoring, survival prediction, and medical cost estimation. Health insurance is now necessary to cover the costs of medical diagnostic and hospital treatment due to the growing cost of healthcare. Consequently, for efficient healthcare planning, it is crucial to precisely forecast individual insurance costs according to demographics, health, lifestyle, and family-related aspects. Existing methods often provide limited prediction accuracy and may not adequately capture complex relationships among these factors. The proposed framework considered a Medical Insurance Cost Dataset containing 1,338 records and 7 attributes and applied data preprocessing, label encoding, Min-Max normalization, SMOTE balancing, and a 70:30 train–test split. Artificial Neural Network (ANN) and Gated Recurrent Unit (GRU) models were developed for medical insurance cost prediction. The proposed GRU achieved an R² of 98.91%, MAE of 222.9553, RMSE of 865.2828, and MAPE of 1.0952, while the ANN achieved an R² of 97.94%, MAE of 508.2767, RMSE of 1156.8298, and MAPE of 2.1902. The results show that the suggested DL framework is capable of accurate and dependable healthcare insurance cost prediction.