Pith. sign in

REVIEW 4 major objections 4 minor 72 references

From Point to probabilistic gradient boosting for claim frequency and severity prediction

T0 review · 4 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Probabilistic gradient boosting can match point-prediction accuracy while improving model adequacy, so insurers do not have to trade off the two.

desk verdict Useful benchmark of ten GBDT algorithms for insurance data, but the abstract's 'no trade-off' claim outruns the evidence—the body's own NB2 and lognormal results show real trade-offs. read the letter →

arxiv 2412.14916 v2 pith:KDI56UNC submitted 2024-12-19 stat.ML cs.LG

Dominik Chevalier , Marie-Pier Côté This is my paper · ORCID
classification stat.MLcs.LG MSC 62P0562-08
keywords gradientboostingdecisiontreesprobabilisticpredictionmodeladequacyCRPSclaimfrequencyseverityinsurancepricing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether actuarial claim frequency and severity models can get the benefits of probabilistic prediction -- a full predictive distribution per policy, not just a point forecast -- without sacrificing the predictive accuracy that makes gradient boosting attractive. It reviews the point and probabilistic GBDT algorithms in a single notation and benchmarks them on five public insurance datasets. Its central finding is that model adequacy and predictive accuracy do not trade off: probabilistic algorithms match point algorithms on deviance-based pseudo-$R^2$, while scoring better on CRPS and confidence-interval coverage, most clearly for lognormal severity. It also reports that LightGBM is the most efficient point model, XGBoostLSS the most efficient probabilistic model, CatBoost the strongest with high-cardinality categoricals, and EGBM the interpretable option at competitive accuracy.

What carries the argument

The central object is the multi-parameter probabilistic GBDT: instead of a single tree sequence for the conditional mean, the algorithm fits $\kappa$ sequences of decision trees, one for each parameter of an assumed distribution (e.g., location and scale for lognormal, mean and dispersion for negative binomial), each with its own link function $\varphi$ to keep predictions in the valid domain, and minimises the negative log-likelihood. XGBoostLSS cycles Newton steps over parameters, NGBoost uses the natural gradient to make updates invariant to reparametrisation, and cyc-GBM uses cyclic coordinate descent; the paper's unified notation is what lets the algorithms be compared fairly. Varying exposure-to-risk is incorporated in frequency models through an offset $\ln e_i$ in the initial prediction, and model adequacy is operationalised with CRPS, DPIT residuals, and confidence-interval coverage.

What would settle it

Re-run every algorithm on 50 random 85/15 splits of the same five datasets and record the pseudo-$R^2$ and CRPS average ranks; if the point-versus-probabilistic ordering flips across splits or the adequacy gaps shrink to within the noise of the splits, the no-trade-off claim would not hold.

Watch

Extended reading notes

Core claim

The central claim is that there is no trade-off between model adequacy and predictive accuracy in gradient boosting for claims. On the paper's own terms, a practitioner can switch from point-prediction GBDTs to probabilistic GBDTs and obtain predictive distributions whose CRPS and confidence-interval coverage are better or equal, while deviance-based McFadden pseudo-$R^2$ stays essentially unchanged. The evidence is the average ranks across five datasets for Poisson and negative-binomial frequency models and lognormal and gamma severity models, with probabilistic methods winning the adequacy rankings in the lognormal case and point methods holding their ground in predictive performance. The paper presents this as a reconciliation of the 'data modelling' and 'algorithmic modelling' cultures in actuarial science.

Load-bearing premise

The conclusion that there is no trade-off rests on the assumption that a single random 85/15 split per dataset is enough to rank the algorithms, and that the five chosen datasets represent actuarial claim frequency and severity problems.

Editorial extensions

If this is right

  • Insurers can adopt probabilistic GBDTs such as XGBoostLSS without expecting a meaningful loss in point predictive performance.
  • Better-calibrated predictive distributions become available for risk management, since CRPS and confidence-interval coverage improve without degrading pseudo-$R^2$.
  • CatBoost's edge on datasets with high-cardinality categorical variables makes it a sensible default when categorical features dominate.
  • EGBM's competitive accuracy makes it a viable interpretable alternative for regulated or stakeholder-facing tariffs.
  • Exposure-to-risk can be handled in tree boosting by the same offset device used in GLMs, so frequency modelling with GBDTs need not abandon that actuarial convention.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper: the single-split design means the no-trade-off claim would be stronger if it survives repeated resampling; the reported average ranks have no stated uncertainty.
  • Beyond the paper: the pattern suggests probabilistic methods will help most where the point loss is location-only, as in the lognormal case, because a gamma deviance already modifies the shape of the fitted distribution.
  • Beyond the paper: combining EGBM's interpretability with probabilistic multi-parameter prediction appears to be an untested but natural next step for actuarial use.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper surveys point and probabilistic gradient boosting for decision trees (GBM, XGBoost, DART, LightGBM, CatBoost, EGBM, PGBM, XGBoostLSS, cyc-GBM, NGBoost) in a unified notation, and compares them on five public claim-frequency and five claim-severity datasets. The numerical study evaluates computational efficiency, McFadden pseudo-R2, CRPS, confidence-interval coverage, portfolio balance, auto-calibration, and tariff segmentation. The central claim, stated in the abstract and again in Section 6, is that there is no trade-off between model adequacy and predictive accuracy: both are achievable simultaneously.

Significance. If the central claim were correct and appropriately scoped, the paper would give practitioners a clear reason to prefer probabilistic boosting when calibrated predictive distributions matter. The unified survey of ten GBDT variants is a useful reference, and the benchmark design has real strengths: separate validation and test splits, identical data partitions across algorithms, five datasets of varied size and cardinality, and multiple evaluation metrics including proper scoring rules and CI coverage. The paper also makes a useful practical point about treating exposure offsets in frequency boosting. However, the headline no-trade-off claim is not supported by the paper's own tables, and the single-split design leaves the key comparative statements without uncertainty quantification. The manuscript is valuable as a comparative study, but its central claim needs to be substantially qualified and re-tested before it can be accepted as stated.

major comments (4)
  1. [Abstract and Section 5.3 (Tables 9 and 16)] The unqualified claim that 'there is no trade-off between model adequacy and predictive accuracy' is contradicted by the NB2 frequency results. Table 9 gives average CRPS ranks of 8.2, 8.4, 9.2, and 7.4 for XGBoostLSS, NGBoost, cyc-GBM, and XGBoostLSSd, versus 2.6, 2.8, and 3.6 for LightGBM, XGBoost, and CatBoost; Table 16 shows the same ordering on all five datasets. The body itself acknowledges in Section 5.3 that 'allowing the shape parameter phi_i to vary with x_i hinders model adequacy, at least on all the studied datasets.' The abstract and Section 6 must be rewritten to state a scoped version of the claim, e.g., one that specifies the distributional settings and algorithms for which adequacy and accuracy are simultaneously achieved, and that explicitly records the NB2 exception.
  2. [Section 5.2 and 5.3 (Tables 7 and 19)] The lognormal BelgianMTPL results are a second direct counterexample to the no-trade-off claim. Table 7 reports McFadden pseudo-R2 values of -0.33 for XGBoostLSS and -0.50 for cyc-GBM, worse than the constant-mean benchmark (GLM 0.27, XGBoost 0.51), while Table 19 shows that these same probabilistic models have the best CI coverage among the GBDT models (e.g., XGBoostLSS coverage 27.90%, 65.62%, 95.33% at the 50%, 75%, 95% levels). This is precisely a trade-off between adequacy and predictive accuracy. The paper needs to either drop the global no-trade-off claim or demonstrate, with additional evidence, why these negative pseudo-R2 values should not count as hindered predictive performance.
  3. [Appendix A (Algorithm 7) and Section 5.2 (Tables 5-9)] The comparative conclusions rest on a single random 85/15 split per dataset, with no repeated runs, no standard errors, and no paired significance tests. Algorithm 7 samples one Dtrain,k and one Dtest,k per dataset, and the average ranks in Tables 5-9 are point estimates. Many of the reported differences are small; for example, Table 5 shows Poisson freMTPL pseudo-R2 of 24.92 for LightGBM and 24.84 for XGBoostLSS, and Table 8 shows WorkComp gamma pseudo-R2 of 33.75 for NGBoost and 32.14 for XGBoost. Without uncertainty information, the statement that probabilistic algorithms do not 'hinder' predictive performance is not established. I recommend adding repeated splits or bootstrap/paired tests, or at minimum reporting standard errors over several seeds, before the comparative claims can be taken as quantitative.
  4. [Section 6 (Discussion)] The discussion paragraph states that 'model adequacy may be enhanced by using probabilistic GBDT without hurting much predictive performance.' This weaker statement is closer to the data, but it still needs a scope qualifier. The paper's own results show that in the NB2 setting the probabilistic algorithms hurt adequacy, and in the lognormal BelgianMTPL setting the best-coverage models have negative pseudo-R2. The discussion should be revised to summarize which algorithms and distributions support the 'no trade-off' reading, rather than presenting it as a general finding.
minor comments (4)
  1. [Table 17] The column header 'Avg. ranl' appears to be a typo for 'Avg. rank'; please fix.
  2. [Section 5.2] The text says 'cyc-GBM looses' where 'loses' is intended; please correct the typo.
  3. [Table 10] The entry for XGBoostLSSd on freMPL reads '2 .27' with an extra space; please correct the formatting.
  4. [General] The paper does not state whether code and trained-model configurations are publicly available. Since Algorithm 7 defines the exact split procedure, a code repository would substantially help reproducibility; please add a data/code availability statement.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; the paper is a benchmark comparison of published algorithms on external public datasets, and its headline claim is an empirical summary rather than a derived or construction-forced result.

full rationale

This paper is a comparative numerical study, not a derivation, so the circularity patterns do not apply in a load-bearing way. The central assertion, 'We find that there is no trade-off between model adequacy and predictive accuracy: both are achievable simultaneously,' is presented as an interpretation of the benchmark tables (Tables 5-9 and 15-18). It is not obtained by defining model adequacy in terms of predictive accuracy, nor by fitting a parameter and then 'predicting' a quantity that is equal to that fit by construction. The predictive performance metric (McFadden pseudo-R2, Eq. 4) and the adequacy metrics (CRPS, DPIT residuals, CI coverage, Section 4.2) are distinct, standard, externally defined scoring rules. All metrics are computed on a held-out 15% test set that is never used for tuning or training (Algorithm 7), so the comparison is not statistically forced by the evaluation setup. The authors' own prior work appears in references [3], [19], and [66] (Henckaerts, Cote, et al.), but those citations are contextual: they support statements about boosting in tariff plans, a hierarchical claims model, and a data preprocessing choice. No benchmark conclusion depends on a theorem, fitted value, or 'uniqueness' claim imported from those self-citations. There is also no ansatz smuggled in via citation and no known result merely renamed in new coordinates. The abstract's unqualified 'no trade-off' sentence is internally inconsistent with some of the paper's own results (e.g., NB2 average CRPS ranks in Table 9 and the negative pseudo-R2 values for lognormal BelgianMTPL in Table 7), but an overbroad empirical claim is a correctness or threat-to-validity concern, not circularity. Similarly, the use of one 85/15 split per dataset without repeated runs weakens the statistical force of the rankings, but it does not make the rankings circular. I find no step in the paper that reduces, by the paper's own equations or by a self-citation chain, to its own inputs.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new entities and no new theory. Its conclusions rest on standard boosting theory, distributional assumptions, evaluation metrics, and hand-chosen hyperparameters. The main non-math assumption is that the chosen datasets and the single split represent actuarial practice well enough to support broad conclusions.

free parameters (3)
  • learning rate lambda = 0.01
    Fixed by hand for all algorithms in Section 4.2 and Appendix A; affects convergence and computational time, so the efficiency comparison depends on it.
  • subsampling proportion delta = 0.75
    Fixed by hand for all algorithms; used in Algorithm 1 and Algorithm 2, affecting training time and stability.
  • minimum samples per leaf = 1% of training observations
    Set in Appendix A; affects tree structure and predictive performance across all compared algorithms.
assumptions (5)
  • standard math Deviance losses (Poisson, gamma) and CRPS are proper scoring rules appropriate for comparing predictive performance and adequacy.
    Used in Section 4.2 to define pseudo-R2 and CRPS; the validity of these metrics is established in the cited literature, not proved here.
  • domain assumption The unified notation in Section 3 faithfully represents the actual implementations of GBM, XGBoost, LightGBM, CatBoost, EGBM, XGBoostLSS, cyc-GBM, NGBoost, and PGBM.
    The comparison's validity depends on correct transcription of each algorithm; errors here would bias the rankings.
  • domain assumption The distributional families (Poisson, negative binomial, lognormal, gamma) are appropriate for the selected datasets.
    Model adequacy (CRPS, CI coverage, DPIT) is evaluated under these assumed distributions; misspecification would make adequacy comparisons hard to interpret.
  • ad hoc to paper A single random 85/15 train/test split per dataset, with tuning on an inner validation split, yields stable enough rankings for the paper's conclusions.
    Algorithm 7 uses one split and no repeated seeds; the no-trade-off conclusion in Sections 5 and 6 assumes these rankings are not noise.
  • standard math The auto-calibration test of Denuit et al. (2024) is correctly implemented.
    The calibration p-values in Tables 11 to 13 rely on this test; improper implementation would change the calibration conclusions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Point to probabilistic gradient boosting for claim frequency and severity prediction." pith.science (2026). https://pith.science/paper/KDI56UNC

@misc{pith2026241214916,
  author       = {Pith},
  title        = {Pith review of: From Point to probabilistic gradient boosting for claim frequency and severity prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KDI56UNC}},
  note         = {Machine review of arXiv:2412.14916}
}
read the original abstract

Gradient boosting for decision tree algorithms are increasingly used in actuarial applications as they show superior predictive performance over traditional generalised linear models. Many enhancements to the first gradient boosting machine algorithm exist. We present in a unified notation, and contrast, all the existing point and probabilistic gradient boosting for decision tree algorithms: GBM, XGBoost, DART, LightGBM, CatBoost, EGBM, PGBM, XGBoostLSS, cyclic GBM, and NGBoost. In this comprehensive numerical study, we compare their performance on five publicly available datasets for claim frequency and severity, of various sizes and comprising different numbers of (high cardinality) categorical variables. We explain how varying exposure-to-risk can be handled with boosting in frequency models. We compare the algorithms on the basis of computational efficiency, predictive performance, and model adequacy. LightGBM and XGBoostLSS win in terms of computational efficiency. CatBoost sometimes improves predictive performance, especially in the presence of high cardinality categorical variables, common in actuarial science. The fully interpretable EGBM achieves competitive predictive performance compared to the black box algorithms considered. We find that there is no trade-off between model adequacy and predictive accuracy: both are achievable simultaneously.

Figures

Figures reproduced from arXiv: 2412.14916 by the authors.

Figure 1
Figure 1. Comprehensive concept map of point (white) and probabilistic (grey) GBDT algorithms. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Predicted lognormal densities given x1 by three algorithms (color and line type), corresponding 95% confidence intervals, and target y1 (black bold vertical line). To measure the adequacy of continuous distributions to severity data, we also consider the coverage of confidence intervals (CIs) on the test set. The probabilistic forecast of an observation leads to an estimated distribution function from which we can c… view at source ↗
Figure 3
Figure 3. Coverage of CI with levels 50%, 75%, and 95% on the test BelgianMTPL set for point (light) or [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Calibration curves of the Poisson models on the test sets. In each bin formed from predicted [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]
Figure 5
Figure 5. Figure 5: Calibration curves of the lognormal models on the test sets. In each bin formed from predicted [PITH_FULL_IMAGE:figures/full_fig_p021_5.png]
Figure 6
Figure 6. Figure 6: Calibration curves of the gamma models on the test sets. In each bin formed from predicted [PITH_FULL_IMAGE:figures/full_fig_p022_6.png]
Figure 7
Figure 7. Figure 7: Main effect for the variable ageph in the Poisson EGBM on BelgianMTPL. 0.100 0.125 0.150 0.175 0.200 20 40 60 80 ageph Partial dependence Model CatBoost cyc.GBM EGBM GBM LightGBM NGBoost XGBoost XGBoostLSS XGBoostLSSd [PITH_FULL_IMAGE:figures/full_fig_p024_7.png]
Figure 8
Figure 8. Figure 8: Partial dependence plots for the variable [PITH_FULL_IMAGE:figures/full_fig_p024_8.png]
Figure 9
Figure 9. Figure 9: Individual conditional expectation plots for the EGBM and CatBoost Poisson models on Belgian [PITH_FULL_IMAGE:figures/full_fig_p025_9.png]
Figure 10
Figure 10. Figure 10: Murphy diagrams comparing the predictive dominance of the Poisson models on the test sets for [PITH_FULL_IMAGE:figures/full_fig_p031_10.png]
Figure 11
Figure 11. Figure 11: Murphy diagrams comparing the predictive dominance of the gamma models on the test sets for [PITH_FULL_IMAGE:figures/full_fig_p032_11.png]
Figure 12
Figure 12. Figure 12: Murphy diagrams comparing the predictive dominance of the lognormal models on the test sets [PITH_FULL_IMAGE:figures/full_fig_p033_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

72 extracted references · 43 canonical work pages

  1. [1]

    Cambridge University Press, Cambridge (2008)

    De Jong, P., Heller, G.Z.: Generalized Linear Models for Insurance Data. Cambridge University Press, Cambridge (2008)

    work page 2008

  2. [2]

    Annual Review of Financial Economics 7(1), 253–277 (2015) https://doi.org/10.1146/annurev-financial-111914-041815

    Frees, E.W.: Analytics of insurance markets. Annual Review of Financial Economics 7(1), 253–277 (2015) https://doi.org/10.1146/annurev-financial-111914-041815

    work page doi:10.1146/annurev-financial-111914-041815 2015

  3. [4]

    Risks 11(9), 163 (2023) https://doi.org/10.3390/risks11090163

    Clemente, C., Guerreiro, G.R., Bravo, J.M.: Modelling motor insurance claim frequency and severity using gradient boosting. Risks 11(9), 163 (2023) https://doi.org/10.3390/risks11090163

    doi:10.3390/risks11090163 2023

  4. [5]

    Statistical Analysis and Data Mining: The ASA Data Science Journal 16(2), 97–119 (2023) https://doi.org/10.1002/sam.11599

    Gao, Y., Huang, Y., Meng, S.: Evaluation and interpretation of driving risks: Automobile claim fre- quency modeling with telematics data. Statistical Analysis and Data Mining: The ASA Data Science Journal 16(2), 97–119 (2023) https://doi.org/10.1002/sam.11599

    work page doi:10.1002/sam.11599 2023

  5. [6]

    Insurance: Mathematics and Economics 104, 158–184 (2022) https://doi.org/10.1016/j

    Crevecoeur, J., Robben, J., Antonio, K.: A hierarchical reserving model for reported non-life insurance claims. Insurance: Mathematics and Economics 104, 158–184 (2022) https://doi.org/10.1016/j. insmatheco.2022.02.005

    doi:10.1016/j 2022

  6. [7]

    North American Actuarial Journal 25(1), 53–61 (2020) https://doi.org/10.1080/10920277.2020.1754242

    Hartman, B., Owen, R., Gibbs, Z.: Predicting high-cost health insurance members through boosted trees and oversampling: An application using the HCCI database. North American Actuarial Journal 25(1), 53–61 (2020) https://doi.org/10.1080/10920277.2020.1754242

    arXiv 2020

  7. [8]

    In: 2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA), pp

    Hancock, J., Khoshgoftaar, T.M.: Performance of CatBoost and XGBoost in medicare fraud detection. In: 2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA), pp. 572–579 (2020). https://doi.org/10.1109/ICMLA51294.2020.00095

    arXiv 2020

  8. [9]

    Annals of statistics 29, 1189–1232 (2001)

    Friedman, J.H.: Greedy function approximation: a gradient boosting machine. Annals of statistics 29, 1189–1232 (2001)

    work page 2001

Show all 72 references
  1. [10]

    Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, 785–794 (2016) https: //doi.org/10.1145/2939672.2939785

    Chen, T., Guestrin, C.: XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, 785–794 (2016) https: //doi.org/10.1145/2939672.2939785

    2016

  2. [11]

    Advances in neural information processing systems 30, 3146– 3154 (2017)

    Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., Liu, T.-Y.: LightGBM: A highly efficient gradient boosting decision tree. Advances in neural information processing systems 30, 3146– 3154 (2017)

    2017

  3. [12]

    Advances in neural information processing systems 31, 6639–6649 (2018)

    Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A.V., Gulin, A.: CatBoost: unbiased boosting with categorical features. Advances in neural information processing systems 31, 6639–6649 (2018)

    2018

  4. [13]

    Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining, 150–158 (2012) https://doi.org/10.1145/2339530.2339556

    Lou, Y., Caruana, R., Gehrke, J.: Intelligible models for classification and regression. Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining, 150–158 (2012) https://doi.org/10.1145/2339530.2339556

    2012

  5. [14]

    Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining, 623–631 (2013) https://doi.org/10.1145/2487575.2487579

    Lou, Y., Caruana, R., Gehrke, J., Hooker, G.: Accurate intelligible models with pairwise interactions. Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining, 623–631 (2013) https://doi.org/10.1145/2487575.2487579

    2013

  6. [15]

    Annual Review of Statistics and Its Application 9, 119–140 (2022) https://doi.org/10.1146/annurev-statistics-040120-030244

    Embrechts, P., W¨ uthrich, M.V.: Recent challenges in actuarial science. Annual Review of Statistics and Its Application 9, 119–140 (2022) https://doi.org/10.1146/annurev-statistics-040120-030244

    2022 doi

  7. [16]

    Insurance: Mathematics and Economics 106, 115–127 (2022) https://doi.org/10.1016/j.insmatheco.2022.06.001

    Meng, S., Gao, Y., Huang, Y.: Actuarial intelligence in auto insurance: Claim frequency modeling with driving behavior features and improved boosted trees. Insurance: Mathematics and Economics 106, 115–127 (2022) https://doi.org/10.1016/j.insmatheco.2022.06.001

    2022 doi

  8. [17]

    Available at SSRN 4352505 (2023) https://doi.org/10.2139/ssrn.4352505

    Delong, L., Lindholm, M., Zakrisson, H.: On cyclic gradient boosting machines. Available at SSRN 4352505 (2023) https://doi.org/10.2139/ssrn.4352505

    2023 doi

  9. [18]

    Available at SSRN 4876585 (2024) https://doi.org/10.48550/arXiv.2406.16206

    So, B., Valdez, E.A.: Zero-inflated Tweedie boosted trees with CatBoost for insurance loss analytics. Available at SSRN 4876585 (2024) https://doi.org/10.48550/arXiv.2406.16206

    2024 doi arXiv

  10. [20]

    Statistical science 16(3), 199–231 (2001) https://doi.org/10.1214/ss/1009213726 26

    Breiman, L.: Statistical modeling: The two cultures (with comments and a rejoinder by the author). Statistical science 16(3), 199–231 (2001) https://doi.org/10.1214/ss/1009213726 26

    2001

  11. [21]

    arXiv preprint arXiv:1907.03178 (2019) https://doi.org/10.48550/arXiv.1907.03178

    M¨ arz, A.: XgboostLSS–an extension of XGBoost to probabilistic forecasting. arXiv preprint arXiv:1907.03178 (2019) https://doi.org/10.48550/arXiv.1907.03178

    2019 doi arXiv

  12. [22]

    Proceedings of the 37th International Conference on Machine Learning 119, 2690–2700 (2020)

    Duan, T., Anand, A., Ding, D.Y., Thai, K.K., Basu, S., Ng, A., Schuler, A.: NGBoost: Natural gradient boosting for probabilistic prediction. Proceedings of the 37th International Conference on Machine Learning 119, 2690–2700 (2020). PMLR

    2020

  13. [23]

    Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 1510–1520 (2021) https://doi.org/10.1145/3447548.3467278

    Sprangers, O., Schelter, S., Rijke, M.: Probabilistic gradient boosting machines for large-scale proba- bilistic regression. Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 1510–1520 (2021) https://doi.org/10.1145/3447548.3467278

    2021

  14. [24]

    Swiss Finance Institute Re- search Paper 16-68 (2023) https://doi.org/10.2139/ssrn.2870308

    W¨ uthrich, M.V., Buser, C.: Data analytics for non-life insurance pricing. Swiss Finance Institute Re- search Paper 16-68 (2023) https://doi.org/10.2139/ssrn.2870308

    2023 doi

  15. [25]

    Springer, Switzerland (2019)

    Denuit, M., Hainaut, D., Trufin, J.: Effective Statistical Learning Methods for Actuaries I: GLMs and Extensions. Springer, Switzerland (2019)

    2019

  16. [26]

    Journal of the American Statistical Association 106(494), 494–510 (2011) https://doi.org/10.1198/jasa.2011.ap09272

    Fenske, N., Kneib, T., Hothorn, T.: Identifying risk factors for severe childhood malnutrition by boosting additive quantile regression. Journal of the American Statistical Association 106(494), 494–510 (2011) https://doi.org/10.1198/jasa.2011.ap09272

    2011 doi

  17. [27]

    Extremes 26(4), 639–667 (2023) https://doi.org/10.1007/s10687-023-00473-x

    Velthoen, J., Dombry, C., Cai, J.-J., Engelke, S.: Gradient boosting for extreme quantile regression. Extremes 26(4), 639–667 (2023) https://doi.org/10.1007/s10687-023-00473-x

    2023 doi

  18. [28]

    Journal of the Royal Statistical Society Series C: Applied Statistics 54(3), 507–554 (2005) https://doi.org/ 10.1111/j.1467-9876.2005.00510.x

    Rigby, R.A., Stasinopoulos, D.M.: Generalized additive models for location, scale and shape. Journal of the Royal Statistical Society Series C: Applied Statistics 54(3), 507–554 (2005) https://doi.org/ 10.1111/j.1467-9876.2005.00510.x

    2005

  19. [30]

    Machine learning 5, 197–227 (1990) https://doi

    Schapire, R.E.: The strength of weak learnability. Machine learning 5, 197–227 (1990) https://doi. org/10.1007/BF00116037

    1990 doi

  20. [31]

    Taylor & Francis, New York (1984)

    Breiman, L., Friedman, J., Olshen, R., Stone, C.: Classification and Regression Trees. Taylor & Francis, New York (1984)

    1984

  21. [32]

    North American Actuarial Journal 22(3), 405–425 (2018) https://doi.org/10.1080/10920277.2018.1431131

    Lee, S.C., Lin, S.: Delta boosting machine with application to general insurance. North American Actuarial Journal 22(3), 405–425 (2018) https://doi.org/10.1080/10920277.2018.1431131

    2018

  22. [33]

    Computational statistics & data analysis 38(4), 367–378 (2002) https://doi.org/10.1016/S0167-9473(01)00065-2

    Friedman, J.H.: Stochastic gradient boosting. Computational statistics & data analysis 38(4), 367–378 (2002) https://doi.org/10.1016/S0167-9473(01)00065-2

    2002 doi

  23. [34]

    Journal of Machine Learning Research 38, 489–497 (2015)

    Rashmi, K.V., Gilad-Bachrach, R.: Dart: Dropouts meet multiple additive regression trees. Journal of Machine Learning Research 38, 489–497 (2015)

    2015

  24. [35]

    Springer, New York (2009)

    Hastie, T., Tibshirani, R., Friedman, J.: The Elements of Statistical Learning: Data Mining, Inference, and Prediction. Springer, New York (2009)

    2009

  25. [36]

    Hickey, J., Metcalfe, P., Ridgeway, G., Schroedl, S., Southworth, H., Therneau, T.: Gbm3: Generalized Boosted Regression Models. (2016). R package version 2.2. https://github.com/gbm-developers/ gbm3

    2016

  26. [37]

    The journal of machine learning research 15(1), 1929–1958 (2014) https://doi.org/10.5555/2627435.2670313

    Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., Salakhutdinov, R.: Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research 15(1), 1929–1958 (2014) https://doi.org/10.5555/2627435.2670313

    2014

  27. [38]

    Machine learning 45, 5–32 (2001) https://doi.org/10.1023/A: 1010933404324 27

    Breiman, L.: Random forests. Machine learning 45, 5–32 (2001) https://doi.org/10.1023/A: 1010933404324 27

    2001 doi

  28. [39]

    Chen, T., He, T., Benesty, M., Khotilovich, V., Tang, Y., Cho, H., Chen, K., Mitchell, R., Cano, I., Zhou, T., Li, M., Xie, J., Lin, M., Geng, Y., Li, Y., Yuan, J.: XGBoost: Extreme Gradient Boosting. (2023). R package version 1.7.5.1. https://CRAN.R-project.org/package=xgboost

    2023

  29. [40]

    Scandinavian Actuarial Journal 10, 1013–1035 (2024) https://doi

    So, B.: Enhanced gradient boosting for zero-inflated insurance claims and comparative analysis of catboost, xgboost, and lightgbm. Scandinavian Actuarial Journal 10, 1013–1035 (2024) https://doi. org/10.1080/03461238.2024.2365390

    2024

  30. [41]

    Shi, Y., Ke, G., Soukhavong, D., Lamb, J., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., Liu, T.-Y., Titov, N.: LightGBM: Light Gradient Boosting Machine. (2023). R package version 3.3.5. https://CRAN.R-project.org/package=lightgbm

    2023

  31. [42]

    R package version 1.2.2 (2023)

    Dorogush, A.V., Ershov, V., Gulin, A.: CatBoost: gradient boosting with categorical features support. R package version 1.2.2 (2023). https://github.com/catboost/catboost

    2023

  32. [43]

    Expert Systems With Ap- plications 167, 114080 (2021) https://doi.org/10.1016/j.eswa.2020.114080

    Sigrist, F.: Gradient and Newton boosting for classification and regression. Expert Systems With Ap- plications 167, 114080 (2021) https://doi.org/10.1016/j.eswa.2020.114080

    2021

  33. [44]

    Official Journal of the European Union L 119, 1–88 (2016)

    GDPR: Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data (general data protection regulation). Official Journal of the...

    2016

  34. [45]

    arXiv preprint arXiv:2404.18702 (2024) https://doi.org/ 10.48550/arXiv.2404.18702

    Xin, X., Huang, F., Hooker, G.: Why you should not trust interpretations in machine learning: adver- sarial attacks on partial dependence plots. arXiv preprint arXiv:2404.18702 (2024) https://doi.org/ 10.48550/arXiv.2404.18702

    2024 doi arXiv

  35. [46]

    arXiv preprint arXiv:1909.09223 (2019) https://doi.org/10.48550/arXiv.1909

    Nori, H., Jenkins, S., Koch, P., Caruana, R.: Interpretml: A unified framework for machine learn- ing interpretability. arXiv preprint arXiv:1909.09223 (2019) https://doi.org/10.48550/arXiv.1909. 09223

    2019 doi arXiv

  36. [47]

    Journal of the Royal Statistical Society Series B: Statistical Methodology 65(1), 95–114 (2003) https://doi.org/10.1111/1467-9868.00374

    Wood, S.N.: Thin plate regression splines. Journal of the Royal Statistical Society Series B: Statistical Methodology 65(1), 95–114 (2003) https://doi.org/10.1111/1467-9868.00374

    2003

  37. [48]

    Journal of Statistical Software 74(1), 1–31 (2016) https://doi.org/10

    Hofner, B., Mayr, A., Schmid, M.: gamboostlss: An r package for model building and variable selection in the gamlss framework. Journal of Statistical Software 74(1), 1–31 (2016) https://doi.org/10. 18637/jss.v074.i01

    2016

  38. [49]

    International series in operations research & management science

    Luenberger, D.G., Ye, Y.: Linear and Nonlinear Programming, Fifth edition edn. International series in operations research & management science. Springer, Cham (2021). https://doi.org/10.1007/ 978-3-030-85450-8

    2021

  39. [50]

    arXiv preprint arXiv:2001.02121 (2020) https://doi.org/10.48550/arXiv.2001.02121

    M¨ arz, A.: CatBoostLSS—an extension of CatBoost to probabilistic forecasting. arXiv preprint arXiv:2001.02121 (2020) https://doi.org/10.48550/arXiv.2001.02121

    2020 doi arXiv

  40. [51]

    https://github

    M¨ arz, A.: LightGBMLSS: An Extension of LightGBM to probabilistic modelling. https://github. com/StatMixedML/LightGBMLSS. GitHub repository, Version 0.4.0 (2023)

    2023

  41. [52]

    Advanced Engineering Informatics 46, 101201 (2020) https://doi.org/10.1016/j.aei.2020.101201

    Chakraborty, D., Elhegazy, H., Elzarka, H., Gutierrez, L.: A novel construction cost prediction model using hybrid natural and light gradient boosting. Advanced Engineering Informatics 46, 101201 (2020) https://doi.org/10.1016/j.aei.2020.101201

    2020

  42. [53]

    Insurance: Mathematics and Economics 101, 485–497 (2021) https://doi

    Denuit, M., Charpentier, A., Trufin, J.: Autocalibration and Tweedie-dominance for insurance pricing with machine learning. Insurance: Mathematics and Economics 101, 485–497 (2021) https://doi. org/10.1016/j.insmatheco.2021.09.001

    2021 doi

  43. [54]

    Frontier in Econometrics 10, 105–142 (1974) 28

    McFadden, D.: Conditional logit analysis of qualitative choice behavior. Frontier in Econometrics 10, 105–142 (1974) 28

    1974

  44. [55]

    Journal of the Royal Statistical Society Series B: Statistical Methodology 78(3), 505–562 (2016) https://doi.org/10.1111/rssb.12154

    Ehm, W., Gneiting, T., Jordan, A., Kr¨ uger, F.: Of quantiles and expectiles: consistent scoring func- tions, choquet representations and forecast rankings. Journal of the Royal Statistical Society Series B: Statistical Methodology 78(3), 505–562 (2016) https://doi.org/10.1111...

    2016 doi

  45. [56]

    arXiv preprint arXiv:2202.12780 (2023) https://doi.org/10.48550/arXiv.2202.12780

    Fissler, T., Lorentzen, C., Mayer, M.: Model comparison and calibration assessment: User guide for consistent scoring functions in machine learning and actuarial practice. arXiv preprint arXiv:2202.12780 (2023) https://doi.org/10.48550/arXiv.2202.12780

    2023 doi arXiv

  46. [57]

    North American Actuarial Journal, 1–44 (2025) https://doi.org/10.1080/10920277.2025.2451860

    Holvoet, F., Antonio, K., Henckaerts, R.: Neural networks for insurance pricing with frequency and severity data: a benchmark study from data preprocessing to technical tariff. North American Actuarial Journal, 1–44 (2025) https://doi.org/10.1080/10920277.2025.2451860

    2025

  47. [58]

    Monthly Weather Review 133(5), 1098–1118 (2005) https://doi.org/10.1175/MWR2904.1

    Gneiting, T., Raftery, A.E., Westveld, A.H., Goldman, T.: Calibrated probabilistic forecasting using ensemble model output statistics and minimum crps estimation. Monthly Weather Review 133(5), 1098–1118 (2005) https://doi.org/10.1175/MWR2904.1

    2005 doi

  48. [59]

    Journal of Statistical Software 90, 1–37 (2019) https://doi.org/10.18637/jss.v090.i12

    Jordan, A., Kr¨ uger, F., Lerch, S.: Evaluating probabilistic forecasts with scoringRules. Journal of Statistical Software 90, 1–37 (2019) https://doi.org/10.18637/jss.v090.i12

    2019 doi

  49. [60]

    Journal of Computational and Graphical Statistics 33, 787–803 (2024) https://doi.org/10.1080/ 10618600.2024.2303336

    Yang, L.: Double probability integral transform residuals for regression models with discrete outcomes. Journal of Computational and Graphical Statistics 33, 787–803 (2024) https://doi.org/10.1080/ 10618600.2024.2303336

    2024

  50. [61]

    Insurance: Mathematics and Economics 117, 130–139 (2024) https://doi.org/10

    Denuit, M., Huyghe, J., Trufin, J., Verdebout, T.: Testing for auto-calibration with lorenz and con- centration curves. Insurance: Mathematics and Economics 117, 130–139 (2024) https://doi.org/10. 1016/j.insmatheco.2024.04.003

    2024

  51. [62]

    Dutang, C., Charpentier, A.: CASdatasets: Insurance Datasets. (2020). R package version 1.0-11

    2020

  52. [63]

    Insurance: Mathematics and Eco- nomics 35(3), 627–647 (2004) https://doi.org/10.1016/j.insmatheco.2004.08.001

    Denuit, M., Lang, S.: Non-life rate-making with Bayesian GAMs. Insurance: Mathematics and Eco- nomics 35(3), 627–647 (2004) https://doi.org/10.1016/j.insmatheco.2004.08.001

    2004 doi

  53. [64]

    Emcien Patterns: Automobile insurance claims including location, policy type and claim amount. (2017). https://emcien.com/sample-data-sets-2/ Accessed 2023-11-21

    2017

  54. [65]

    Priest, C.: Predict workers compensation claims using highly realistic synthetic data. (2021). https:// www.kaggle.com/datasets/lucamassaron/easy-peasy-its-lemon-squeezy/ Accessed 2023-11-14

    2021

  55. [66]

    Expert Systems with Applications 202, 117230 (2022) https://doi.org/10.1016/j.eswa.2022.117230

    Henckaerts, R., Antonio, K., Cˆ ot´ e, M.-P.: When stakes are high: Balancing accuracy and trans- parency with model-agnostic interpretable data-driven surrogates. Expert Systems with Applications 202, 117230 (2022) https://doi.org/10.1016/j.eswa.2022.117230

    2022

  56. [67]

    Ohlsson, E., Johansson, B.: Non-life Insurance Pricing with Generalized Linear Models vol. 174. Springer, New York (2010)

    2010

  57. [68]

    North American Actuarial Journal 28(2), 285–319 (2024) https://doi.org/10.1080/10920277.2023

    Xin, X., Huang, F.: Antidiscrimination insurance pricing: Regulations, fairness criteria, and models. North American Actuarial Journal 28(2), 285–319 (2024) https://doi.org/10.1080/10920277.2023. 2190528

    2024

  58. [69]

    Springer, Cham (2023)

    W¨ uthrich, M.V., Merz, M.: Statistical Foundations of Actuarial Learning and Its Applications. Springer, Cham (2023)

    2023

  59. [70]

    Journal of Risk and Insurance 81(2), 335–366 (2014) https://doi.org/10.1111/j.1539-6975.2012.01507.x

    Frees, E.W., Meyers, G., Cummings, A.D.: Insurance ratemaking and a Gini index. Journal of Risk and Insurance 81(2), 335–366 (2014) https://doi.org/10.1111/j.1539-6975.2012.01507.x

    2014

  60. [71]

    Klein, N., Denuit, M., Lang, S., Kneib, T.: Nonlife ratemaking and risk management with Bayesian generalized additive models for location, scale, and shape. Insurance: Mathematics and Economics 55, 225–249 (2014) https://doi.org/10.1016/j.insmatheco.2014.02.001 29 Algorithm 7:...

    2014 doi

  61. [72]

    journal of Computational and Graphical Statistics 24(1), 44–65 (2015) https://doi.org/10.1080/10618600.2014.907095

    Goldstein, A., Kapelner, A., Bleich, J., Pitkin, E.: Peeking inside the black box: Visualizing statisti- cal learning with plots of individual conditional expectation. journal of Computational and Graphical Statistics 24(1), 44–65 (2015) https://doi.org/10.1080/10618600.2014.907095

    2015

  62. [73]

    The Annals of Applied Statistics 2(3), 916–954 (2008) https://doi.org/10.1214/07-AOAS148

    Friedman, J.H., Popescu, B.E.: Predictive learning via rule ensembles. The Annals of Applied Statistics 2(3), 916–954 (2008) https://doi.org/10.1214/07-AOAS148

    2008 doi

  63. [74]

    European Actuarial Journal 10(1), 179–202 (2020) https://doi.org/10.1007/s13385-019-00215-z

    W¨ uthrich, M.V.: Bias regularization in neural network models for general insurance pricing. European Actuarial Journal 10(1), 179–202 (2020) https://doi.org/10.1007/s13385-019-00215-z

    2020 doi

  64. [75]

    W¨ uthrich, M.V., Ziegel, J.: Isotonic recalibration under a low signal-to-noise ratio. Scandinavian Actu- arial Journal 2024(3), 279–299 (2024) https://doi.org/10.1080/03461238.2023.2246743 A Tuning strategy We follow Algorithm 7 to tune hyperparameters with the five datasets...

    2024

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.