Abstract
Health insurance ensures access to care, medical cost containment, and improved health outcomes. It also promotes preventive care and ensures financial protection, thereby contributing to individual and societal well-being. In many countries, health insurance has been legislated. Within such frameworks, insurers are required to design customized plans for specific risk profiles by applying machine learning (ML) models to estimate insurance premiums based on factors like age, region, and coverage level. This study employs Multi-Layer Perceptron Regression (MLPR) to identify underlying patterns and generate personalized premium estimates. For this reason, the MLPR model architecture includes an input layer, one or more hidden layers, and an output layer, which collectively enable the model to capture complex nonlinear relationships inherent in the data. Besides, metaheuristic optimization algorithms, including the Aquila Optimizer (AO) and the Turbulent Flow of Water-based Optimization (TFWO), were applied to enhance model reliability and predictive performance. Among the analyzed third-layer models, the MLTF model demonstrated the strongest performance, achieving R2 value of 0.989 and an RMSE of 2765. This was followed by the MLAO model at the third layer, which attained R2 value of 0.974 and an RMSE of 3517, indicating comparable predictive accuracy.
Recommended Citation
Meng, Jie and Meng, Jing
(2026)
"Health Insurance Rate Optimal Predictions Through Auto-Tuning Methods,"
Iraqi Journal for Computer Science and Mathematics: Vol. 7:
Iss.
4, Article 1.
DOI: https://doi.org/10.52866/2788-7421.1424
Available at:
https://ijcsm.researchcommons.org/ijcsm/vol7/iss4/1

