Reseach Article
Dynamic Car Insurance Pricing using Interpretable Machine Learning Models: A Comparative Study of Regression, Support Vector Machines, and Neural Networks
| Journal of Advanced Artificial Intelligence |
| Foundation of Computer Science (FCS), NY, USA |
| Volume 2 - Number 4 |
| Year of Publication: 2026 |
| Authors: Yassmine Ouladhaj Kaddour, Asmaa Faris, Mohamed Dakkoun |
10.5120/jaai202659
|
Yassmine Ouladhaj Kaddour, Asmaa Faris, Mohamed Dakkoun . Dynamic Car Insurance Pricing using Interpretable Machine Learning Models: A Comparative Study of Regression, Support Vector Machines, and Neural Networks. Journal of Advanced Artificial Intelligence. 2, 4 ( Jan 2026), 1-12. DOI=10.5120/jaai202659
Abstract
In the non-life insurance sector, dynamic pricing has emerged as a crucial component of modern auto insurance, allowing insurers to adjust premiums more precisely and enhance risk differentiation. Traditional actuarial methods, such as generalized linear models (GLMs), provide strong interpretability but often fail to capture complex nonlinear relationships or high-dimensional structures. To address these limitations, this study explores supervised learning techniques, including regression models, support vector machines (SVMs), and neural networks (NNs), to model behavioral patterns, driving conditions, and claim outcomes. The results indicate that simple linear models consistently outperform more complex approaches. Ridge, Lasso, and Linear Regression achieve comparable performance, with R2 values around 0.943 and RMSE ranging from 97.56 to 97.71. Interpretability analyses, including permutation importance and SHAP, reveal that prior accidents are the primary determinant of pricing decisions, exerting an effect nearly ten times greater than other factors. The overall objective is to develop an accurate and interpretable model capable of estimating premiums while offering improvements over traditional actuarial methods.
References
- Modugno, L., Perin, G., & Bernardi, A. (2023). Modelling motor insurance claim frequency and severity using gradient boosting. Risks, 11(3).
- Henckaerts, R., Frees, E. W., & Antonio, K. (2023). Neural networks for insurance pricing with frequency and severity data: A benchmark study from data preprocessing to technical tariff. North American Actuarial Journal, 29(3).
- Mahendran, A., Thompson, H., & McGree, J. M. (2023). A model-robust subsampling approach for generalized linear models in big data settings. Journal of Computational and Graphical Statistics, 32(2).
- Rayadurgam, S., & Byun, T. (2020). Manifold for machine learning assurance. In Proceedings of ICSE-NIER 2020.
- Burton, S., & Herd, B. (2023). Addressing uncertainty in the safety assurance of machine learning. IEEE Software.
- Haris, M., & Arum, P. (2022). Poisson-gamma and inverse- Poisson models for insurance claim modeling. BAREKENG: Jurnal Ilmu Matematika dan Terapan, 16(2).
- Novkaniza, A. (2025). A posteriori premium rate calculation using Poisson-gamma hierarchical generalized linear model. Journal of Theoretical and Applied Mathematics (JTAM), 12(4).
- W¨uthrich, M. V., & Merz, M. (2023). Statistical foundations of actuarial learning and its applications. Springer Actuarial.
- McGuire, G., Taylor, G., & Miller, H. (2021). Self-assembling insurance claim models using regularized regression and machine learning. Variance, 14(1).
- Kennard, R., & Hoerl, A. (2000). Ridge regression: Biased estimation for nonorthogonal problems. Technometrics, 42(1).
- Sun, B., Lin, X., Furman, E., & W¨uthrich, M. (2024). Nonparametric intercept regularization for insurance claim frequency regression models. ASTIN Bulletin, 54(1).
- Vapnik, V. (2000). The nature of statistical learning theory (2nd ed.). Springer.
- W¨uthrich, M. V. (2019). Neural network applications for actuarial risk modelling. ASTIN Bulletin, 49(3).
- Duval, F., Boucher, J.-P., & Pigeon, M. (2024). Telematics combined actuarial neural networks for cross-sectional and longitudinal claim count data. ASTIN Bulletin.
- Kopalle, P. K., Pauwels, K., Akella, L. Y., & Gangwar, M. (2023). Dynamic pricing: Definition, implications for managers, and future research directions. Journal of the Academy of Marketing Science, 51(1).
- Friedman, J., Hastie, T., & Tibshirani, R. (2010). Regularization paths for generalized linear models via coordinate descent. Journal of Statistical Software, 33(1).
- Hair, J. F., Black, W. C., Babin, B. J., & Anderson, R. E. (2018). Multivariate data analysis (8th ed.). Pearson.
- Haris, M., & Arum, P. (2022). Negative binomial regression and generalized Poisson regression models on the number of traffic accidents in Central Java. BAREKENG: Jurnal Ilmu Matematika dan Terapan, 16(2).
- Hastie, T., Tibshirani, R.,&Friedman, J. (2001). The elements of statistical learning. Springer.
- Kumar, I. E., Venkatasubramanian, S., Scheidegger, C., & Friedler, S. (2022). Problems with Shapley-value-based explanations as feature importance measures. Proceedings of the AAAI Conference on Artificial Intelligence, 36(7).
- Slack, D., Hilgard, S., Jia, E., Singh, S., & Lakkaraju, H. (2022). Fooling LIME and SHAP: Adversarial attacks and robustness improvements. Machine Learning, 111(5).
- Cervantes, F., & Gomez, R. (2020). Support vector machines applications in insurance risk prediction. Journal of Insurance Analytics, 8(1).
Index Terms
Keywords