{"paper_id":"2409.01908","title":"Bayesian CART models for aggregate claim modeling","authors":["Yaojun Zhang","Lanpeng Ji","Georgios Aivaliotis","Charles C. Taylor"],"paper_date":"2024-09-03","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2409.01908v2","page_url":"https://thequant.space/flowcharts/bayesian-cart-models-for-aggregate-claim-modeling/","scores":{"empirical_rigor":7.5,"math_complexity":8,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes three types of Bayesian CART models for aggregate claim amount: frequency‑severity, sequential, and joint models.","quote":"This paper proposes three types of Bayesian CART (or BCART) models for aggregate claim amount, namely, frequency-severity models, sequential models and joint models.","summary":"This paper proposes three types of Bayesian CART (or BCART) models for aggregate claim amount, namely, frequency-severity models, sequential models and joint models. We propose a general framework for the BCART models applicable to data with multivariate responses, which is particularly useful for…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Exposure vi is set to 1 for simplicity in the simulation.","type":"data","stated":"explicit","quote":"We assume exposure vi ≡1 for simplicity."},{"assumption":"Conditional on the tree T, the parameters θt are independent and identically distributed with prior p(θt).","type":"model","stated":"explicit","quote":"we assume that conditional on the tree T with b terminal nodes as above, the parameters θt, t = 1, 2, . . . , b, have IID priors p(θt)"},{"assumption":"Each θt is split into known (θt,M) and unknown (θt,B) components, with the known part estimated by MME or MLE.","type":"model","stated":"explicit","quote":"We assume θt = (θt,M, θt,B), where θt,M are the parameters that are treated as known and computed using Method of Moments Estimation (MME), or Maximum Likelihood Estimation (MLE), and θt,B are the unknown parameters that need to be estimated in the Bayesian framework."}]},"datasets":{"status":"available","items":[{"name":"dataCar","provider":"insuranceData library in R","kind":"other","access":"public","period":null,"universe":"67,856 vehicle insurance policies (2004‑2005)","quote":"We illustrate our methodology with the insurance dataset dataCar, available from the library insuranceData in R; see [53] for details. This dataset is based on one-year vehicle insurance policies taken out in 2004 or 2005. There are 67,856 policies of which 93.19% made no claims."},{"name":"Scenario 1 simulated data","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"with n = 5,000 independent observations. Here xi = (xi1, xi2), with independent components xik ∼N(0, 1) for k = 1, 2. We assume exposure vi ≡1 for simplicity."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"unclear"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R3","candidates":{"front_matter":"n/a","evidence_card":"n/a","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"unclear","findings":[{"finding":"The Weibull distribution outperforms gamma and lognormal distributions for modeling claim severity in BCART trees.","kind":"statistical","quote":"We discover that the Weibull distribution is superior to gamma and lognormal distributions, due to its ability to capture different tail characteristics in tree models."},{"finding":"Sequential and joint BCART models that incorporate dependence between claim count and average severity are preferable to independent frequency‑severity models.","kind":"performance","quote":"we find that sequential BCART models and joint BCART models, which incorporate dependence between the number of claims and average severity, are beneficial and thus preferable to the frequency-severity BCART models in which independence is assumed."},{"finding":"Among joint models, ZICPG3‑BCART achieves the lowest DIC (102120) and is considered the best model.","kind":"performance","quote":"ZICPG3-BCART, with the smallest DIC (= 102120), is deemed to be the best."},{"finding":"Including the claim count N or its estimate ˆN as a covariate improves model performance across all evaluation metrics.","kind":"performance","quote":"adding N (or ˆN) as a covariate improves performance, as shown by all evaluation metrics"},{"finding":"When two trees have similar splitting rules, a joint tree is more effective through information sharing; otherwise, separate trees outperform a joint tree.","kind":"theoretical","quote":"when two trees have similar splitting rules (high ARI), one joint tree is more effective through information sharing. Conversely, if all covariates affecting claim frequency and average severity are different (ARI is close to −1), two trees outperform one joint tree."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":20,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T18:03:07Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}