Leveraging AI for Mortality Risk Prediction in Life Insurance: Techniques, Models, and Real-World Applications
Keywords:
Artificial Intelligence, Machine LearningAbstract
Accurate assessments of mortality risk are very vital for life insurers to thrive and compete. Medical history and self-reported data-based conventional underwriting cannot sufficiently depict the complicated interactions among lifespan factors. Particularly ML, artificial intelligence has developed to fulfill this requirement. Methodologies, model development, validation, and practical applications to enhance underwriting in life insurance concerning artificial intelligence mortality risk prediction are discussed in this work.
Synopsis: This work investigates accuracy of life insurance underwriting and mortality risk assessment. We next show how traditional approaches cannot adequately explain developing risk variables and judgment biases. Ideas of artificial intelligence and machine learning including supervised learning mortality risk prediction systems are then under investigation. One may compare logistic regression, random forests, and gradient boosting.
References
[1] A. I. Koning and M. J. van Wissen, "Mortality risk assessment in a life insurance context using proportional hazards models," Insurance: Mathematics and Economics, vol. 46, no. 2, pp. 271-280, 2009. [doi: 10.1016/j.insmatheco.2008.08.002]
[2] Y. Luo et al., "Predicting mortality risk using deep neural networks with long short-term memory," Insurance: Mathematics and Economics, vol. 81, pp. 104-111, 2018. [doi: 10.1016/j.insmatheco.2018.02.003]
[3] H. Ding et al., "Explainable machine learning for mortality risk prediction in life insurance," arXiv preprint arXiv:2002.02153, 2020.
[4] I. Witten, E. Frank, and M. A. Hall, Data Mining: Practical Machine Learning Tools and Techniques (Morgan Kaufmann Series in Data Management Systems), Morgan Kaufmann Publishers, 2016.
[5] D. Pyle, Data Science Handbook (2nd ed.), O'Reilly Media, Inc., 2015.
[6] A. Géron, Hands-On Machine Learning with Scikit-Learn, Keras & TensorFlow (Concepts, Tools, and Techniques to Build Intelligent Systems), O'Reilly Media, Inc., 2017.
[7] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning (Springer Series in Statistics), Springer New York Inc., 2009.
[8] G. James, D. Witten, T. Hastie, and R. Tibshirani, An Introduction to Statistical Learning (Springer Publishing Company, Incorporated), 2013.
[9] Y. Bengio, "Tuning hyperparameters of a learning algorithm," Neural Networks, vol. 13, no. 8, pp. 815-828, 2000. [doi: 10.1016/S0893-6080(00)00031-X]
[10] T. G. Dietterich, "Approximate statistical tests for comparing supervised learning algorithms," Artificial intelligence, vol. 85, no. 1-2, pp. 1-31, 1997. [doi: 10.1016/S0004-3702(96)00061-X]
[11] P. Christopoulos et al., "Calibration and refinement of machine learning models for predicting mortality risk," Biostatistics, vol. 21, no. 1, pp. 22-31, 2020. [doi: 10.1093/biostat/fst083]
[12] N. Naeini, G. S. Ongmongkolkul, and J. D. Walker, "Obtaining well-calibrated risk estimates in machine learning models: The importance of calibration," PLoS One, vol. 13, no. 5, e0195159, 2018. [doi: 10.1371/journal.pone.0195159]
[13] S. Mehrabi et al., "A survey of bias in machine learning," arXiv preprint arXiv:1908.11592, 2019.
[14] A. Selbst, D. Boyd, S. Crawford, and M. Gebru, "Fairness in artificial intelligence algorithmic bias, algorithmic fairness, and limits of fairness in machine learning," AI Now Institute, 2019.
