Autocalibration and Tweedie-dominance for insurance pricing with machine learning
From MaRDI portal
Publication:2665871
DOI10.1016/J.INSMATHECO.2021.09.001zbMATH Open1475.91295arXiv2103.03635OpenAlexW3200416260MaRDI QIDQ2665871FDOQ2665871
Authors: Arthur Charpentier, Julien Trufin, Michel Denuit 
Publication date: 19 November 2021
Published in: Insurance Mathematics \& Economics (Search for Journal in Brave)
Abstract: Boosting techniques and neural networks are particularly effective machine learning methods for insurance pricing. Often in practice, there are nevertheless endless debates about the choice of the right loss function to be used to train the machine learning model, as well as about the appropriate metric to assess the performances of competing models. Also, the sum of fitted values can depart from the observed totals to a large extent and this often confuses actuarial analysts. The lack of balance inherent to training models by minimizing deviance outside the familiar GLM with canonical link setting has been empirically documented in W"uthrich (2019, 2020) who attributes it to the early stopping rule in gradient descent methods for model fitting. The present paper aims to further study this phenomenon when learning proceeds by minimizing Tweedie deviance. It is shown that minimizing deviance involves a trade-off between the integral of weighted differences of lower partial moments and the bias measured on a specific scale. Autocalibration is then proposed as a remedy. This new method to correct for bias adds an extra local GLM step to the analysis. Theoretically, it is shown that it implements the autocalibration concept in pure premium calculation and ensures that balance also holds on a local scale, not only at portfolio level as with existing bias-correction techniques. The convex order appears to be the natural tool to compare competing models, putting a new light on the diagnostic graphs and associated metrics proposed by Denuit et al. (2019).
Full work available at URL: https://arxiv.org/abs/2103.03635
Recommendations
- Boosting insights in insurance tariff plans with tree-based machine learning methods
- Bias regularization in neural network models for general insurance pricing
- Response versus gradient boosting trees, GLMs and neural networks under Tweedie loss and log-link
- A Markov-modulated tree-based gradient boosting model for auto-insurance risk premium pricing
- Actuarial intelligence in auto insurance: claim frequency modeling with driving behavior features and improved boosted trees
Cites Work
- Local Regression and Likelihood
- Stochastic orders
- Expectation dependence of random variables, with an application in portfolio theory
- Global dependence stochastic orders
- Bias regularization in neural network models for general insurance pricing
- Making Tweedie's compound Poisson model more accessible
- Model selection based on Lorenz and concentration curves, Gini indices and convex order
Cited In (13)
- Telematics combined actuarial neural networks for cross-sectional and longitudinal claim count data
- Testing for more positive expectation dependence with application to model comparison
- Testing for auto-calibration with Lorenz and concentration curves
- Bayesian CART models for insurance claims frequency
- Bias regularization in neural network models for general insurance pricing
- Convex and Lorenz orders under balance correction in nonlife insurance pricing: review and new developments
- Local bias adjustment, duration-weighted probabilities, and automatic construction of tariff cells
- Model selection with Gini indices under auto-calibration
- Enhancing actuarial non-life pricing models via transformers
- On duration effects in non-life insurance pricing
- Boosting insights in insurance tariff plans with tree-based machine learning methods
- Deep quantile and deep composite triplet regression
- Isotonic recalibration under a low signal-to-noise ratio
Uses Software
This page was built for publication: Autocalibration and Tweedie-dominance for insurance pricing with machine learning
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2665871)
