Applying economic measures to lapse risk management with machine learning approaches

Stéphane Loisel (1) , Pierrick Piette (1, 2) , Jason Tsai (3)

1 SAF - Laboratoire de Sciences Actuarielle et Financière
2 LPSM (UMR_8001) - Laboratoire de Probabilités, Statistique et Modélisation
3 NCCU - National ChengChi University [Taipei]

Modeling policyholders lapse behaviors is important to a life insurer since lapses affect pricing, reserving, profitability, liquidity, risk management, as well as the solvency of the insurer. Lapse risk is indeed the most significant life underwriting risk according to European Insurance and Occupational Pensions Authority's Quantitative Impact Study QIS5. In this paper, we introduce two advanced machine learning algorithms for lapse modeling. Then we evaluate the performance of different algorithms by means of classical statistical accuracy and profitability measure. Moreover, we adopt an innovative point of view on the lapse prediction problem that comes from churn management. We transform the classification problem into a regression question and then perform optimization, which is new for lapse risk management. We apply different algorithms to a large real-world insurance dataset. Our results show that XGBoost and SVM outperform CART and logistic regression, especially in terms of the economic validation metric. The optimization after transformation brings out significant and consistent increases in economic gains.

Mots clés: lapse; machine learning; SVM; XGBoost; life insurance

Domaines: Applications [stat.AP]

Stéphane Loisel, Pierrick Piette, Jason Tsai. Applying economic measures to lapse risk management with machine learning approaches. 2019. ⟨hal-02150983v1⟩