arXiv subjects

Katrien Antonio

Publications and source records attributed to Katrien Antonio.

At least 19 recordsLinked to original sources

arXiv2026

A multi-view contrastive learning framework for spatial embeddings in risk modelling

Incorporating spatial information, particularly when related to climate, weather, and demographic factors, is crucial for improving underwriting precision and enhancing risk management in insurance. However, spatial data are often unstructured, high-dimensional, and difficult to integrate into predictive models. Embedding methods are needed to convert spatial data into meaningful representations for modelling tasks. We propose a novel multi-view contrastive learning framework for generating spatial embeddings that combine information from multiple spatial data sources. To train the model, we construct a spatial dataset that merges satellite imagery and OpenStreetMap features across Europe. The framework aligns these spatial views with coordinate-based encodings, producing low-dimensional embeddings that capture both spatial structure and contextual similarity. Once trained, the model generates embeddings directly from latitude-longitude pairs, enabling any dataset with coordinates to be enriched with meaningful spatial features without requiring access to the original spatial inputs. In a case study on French real estate prices, we compare models trained on raw coordinates against those using our spatial embeddings as inputs. The embeddings consistently improve predictive accuracy across generalised linear, additive, and boosting models, while providing post-hoc explainable spatial effects and demonstrating generalisation of the fitted spatial effects to regions without training observations. A second case study on flood claim counts across Belgian postal codes confirms that the embeddings improve territorial risk classification in an insurance context.

q-fin.RM↗

arXiv2025

Machine learning in an expectation-maximisation framework for nowcasting

Decision making often occurs in the presence of incomplete information, leading to the under- or overestimation of risk. Leveraging the observable information to learn the complete information is called nowcasting. In practice, incomplete information is often a consequence of reporting or observation delays. In this paper, we propose an expectation-maximisation (EM) framework for nowcasting that uses machine learning techniques to model both the occurrence as well as the reporting process of events. We allow for the inclusion of covariate information specific to the occurrence and reporting periods as well as characteristics related to the entity for which events occurred. We demonstrate how the maximisation step and the information flow between EM iterations can be tailored to leverage the predictive power of neural networks and (extreme) gradient boosting machines (XGBoost). With simulation experiments, we show that we can effectively model both the occurrence and reporting of events when dealing with high-dimensional covariate information. In the presence of non-linear effects, we show that our methodology outperforms existing EM-based nowcasting frameworks that use generalised linear models in the maximisation step. Finally, we apply the framework to the reporting of Argentinian Covid-19 cases, where the XGBoost-based approach again is most performant.

stat.ML↗

arXiv2025

Machine Learning with Multitype Protected Attributes: Intersectional Fairness through Regularisation

Ensuring equitable treatment (fairness) across protected attributes (such as gender or ethnicity) is a critical issue in machine learning. Most existing literature focuses on binary classification, but achieving fairness in regression tasks-such as insurance pricing or hiring score assessments-is equally important. Moreover, anti-discrimination laws also apply to continuous attributes, such as age, for which many existing methods are not applicable. In practice, multiple protected attributes can exist simultaneously; however, methods targeting fairness across several attributes often overlook so-called "fairness gerrymandering", thereby ignoring disparities among intersectional subgroups (e.g., African-American women or Hispanic men). In this paper, we propose a distance covariance regularisation framework that mitigates the association between model predictions and protected attributes, in line with the fairness definition of demographic parity, and that captures both linear and nonlinear dependencies. To enhance applicability in the presence of multiple protected attributes, we extend our framework by incorporating two multivariate dependence measures based on distance covariance: the previously proposed joint distance covariance (JdCov) and our novel concatenated distance covariance (CCdCov), which effectively address fairness gerrymandering in both regression and classification tasks involving protected attributes of various types. We discuss and illustrate how to calibrate regularisation strength, including a method based on Jensen-Shannon divergence, which quantifies dissimilarities in prediction distributions across groups. We apply our framework to the COMPAS recidivism dataset and a large motor insurance claims dataset.

cs.LG↗

arXiv2025

Neural networks for insurance pricing with frequency and severity data: a benchmark study from data preprocessing to technical tariff

Insurers usually turn to generalized linear models for modeling claim frequency and severity data. Due to their success in other fields, machine learning techniques are gaining popularity within the actuarial toolbox. Our paper contributes to the literature on frequency-severity insurance pricing with machine learning via deep learning structures. We present a benchmark study on four insurance data sets with frequency and severity targets in the presence of multiple types of input features. We compare in detail the performance of: a generalized linear model on binned input data, a gradient-boosted tree model, a feed-forward neural network (FFNN), and the combined actuarial neural network (CANN). The CANNs combine a baseline prediction established with a GLM and GBM, respectively, with a neural network correction. We explain the data preprocessing steps with specific focus on the multiple types of input features typically present in tabular insurance data sets, such as postal codes, numeric and categorical covariates. Autoencoders are used to embed the categorical variables into the neural network, and we explore their potential advantages in a frequency-severity setting. Model performance is evaluated not only on out-of-sample deviance but also using statistical and calibration performance criteria and managerial tools to get more nuanced insights. Finally, we construct global surrogate models for the neural nets' frequency and severity models. These surrogates enable the translation of the essential insights captured by the FFNNs or CANNs to GLMs. As such, a technical tariff table results that can easily be deployed in practice.

cs.LG↗

arXiv2024

An engine to simulate insurance fraud network data

Traditionally, the detection of fraudulent insurance claims relies on business rules and expert judgement which makes it a time-consuming and expensive process (Óskarsdóttir et al., 2022). Consequently, researchers have been examining ways to develop efficient and accurate analytic strategies to flag suspicious claims. Feeding learning methods with features engineered from the social network of parties involved in a claim is a particularly promising strategy (see for example Van Vlasselaer et al. (2016); Tumminello et al. (2023)). When developing a fraud detection model, however, we are confronted with several challenges. The uncommon nature of fraud, for example, creates a high class imbalance which complicates the development of well performing analytic classification models. In addition, only a small number of claims are investigated and get a label, which results in a large corpus of unlabeled data. Yet another challenge is the lack of publicly available data. This hinders not only the development of new methods, but also the validation of existing techniques. We therefore design a simulation machine that is engineered to create synthetic data with a network structure and available covariates similar to the real life insurance fraud data set analyzed in Óskarsdóttir et al. (2022). Further, the user has control over several data-generating mechanisms. We can specify the total number of policyholders and parties, the desired level of imbalance and the (effect size of the) features in the fraud generating model. As such, the simulation engine enables researchers and practitioners to examine several methodological challenges as well as to test their (development strategy of) insurance fraud detection models in a range of different settings. Moreover, large synthetic data sets can be generated to evaluate the predictive performance of (advanced) machine learning techniques.

cs.LG↗

arXiv2024

The short-term association between environmental variables and mortality: evidence from Europe

Using fine-grained, publicly available data, this paper studies the short-term association between environmental factors, i.e., weather and air pollution characteristics, and weekly mortality rates in small geographical regions in Europe. Hereto, we develop a mortality modeling framework where a baseline model describes a region-specific, seasonal trend observed within the historical weekly mortality rates. Using a machine learning algorithm, we then explain deviations from this baseline using features constructed from environmental data that capture anomalies and extreme events. We illustrate our proposed modeling framework through a case study on more than 550 NUTS 3 regions (Nomenclature of Territorial Units for Statistics, level 3) in 20 European countries. Using interpretation tools, we unravel insights into which environmental features are most important when estimating excess or deficit mortality relative to the baseline and explore how these features interact. Moreover, we investigate harvesting effects through our constructed weekly mortality modeling framework. Our findings show that temperature-related features are most influential in explaining mortality deviations from the baseline over short time periods. Furthermore, we find that environmental features prove particularly beneficial in southern regions for explaining elevated levels of mortality, and we observe evidence of a harvesting effect related to heat waves.

stat.AP↗

arXiv2024

Reducing the dimensionality and granularity in hierarchical categorical variables

Hierarchical categorical variables often exhibit many levels (high granularity) and many classes within each level (high dimensionality). This may cause overfitting and estimation issues when including such covariates in a predictive model. In current literature, a hierarchical covariate is often incorporated via nested random effects. However, this does not facilitate the assumption of classes having the same effect on the response variable. In this paper, we propose a methodology to obtain a reduced representation of a hierarchical categorical variable. We show how entity embedding can be applied in a hierarchical setting. Subsequently, we propose a top-down clustering algorithm which leverages the information encoded in the embeddings to reduce both the within-level dimensionality as well as the overall granularity of the hierarchical categorical variable. In simulation experiments, we show that our methodology can effectively approximate the true underlying structure of a hierarchical covariate in terms of the effect on a response variable, and find that incorporating the reduced hierarchy improves the balance between model fit and complexity. We apply our methodology on a real dataset and find that the reduced hierarchy is an improvement over the original hierarchical structure and reduced structures proposed in the literature.

stat.ME↗

arXiv2024

On clustering levels of a hierarchical categorical risk factor

Handling nominal covariates with a large number of categories is challenging for both statistical and machine learning techniques. This problem is further exacerbated when the nominal variable has a hierarchical structure. We commonly rely on methods such as the random effects approach (Campo and Antonio, 2023) to incorporate these covariates in a predictive model. Nonetheless, in certain situations, even the random effects approach may encounter estimation problems. We propose the data-driven Partitioning Hierarchical Risk-factors Adaptive Top-down (PHiRAT) algorithm to reduce the hierarchically structured risk factor to its essence, by grouping similar categories at each level of the hierarchy. We work top-down and engineer several features to characterize the profile of the categories at a specific level in the hierarchy. In our workers' compensation case study, we characterize the risk profile of an industry via its observed damage rates and claim frequencies. In addition, we use embeddings (Mikolov et al., 2013; Cer et al., 2018) to encode the textual description of the economic activity of the insured company. These features are then used as input in a clustering algorithm to group similar categories. Our method substantially reduces the number of categories and results in a grouping that is generalizable to out-of-sample data. Moreover, we obtain a better differentiation between high-risk and low-risk companies.

stat.ME↗

arXiv2023

Catastrophe Risk in a Stochastic Multi-Population Mortality Model

This paper presents an approach to incorporate mortality shocks into mortality projections produced by a stochastic multi-population mortality model. The proposed model combines a decreasing stochastic mortality trend with a regime-switching mechanism that captures age-specific mortality shocks over a lengthy calibration period. The result is a flexible and powerful toolbox that actuaries and risk managers can tailor to their specific needs, risk appetite, or supervisory requirements. We illustrate the proposed mortality model with a case study on projecting Dutch mortality rates. Our findings show that the proposed model generates wider prediction intervals for the mortality rates compared to state-of-the-art stochastic mortality models. The width of these prediction intervals depends on the frequency and severity of the mortality shocks calibrated with the regime-switching model. Furthermore, we compare the solvency capital requirement (SCR) for mortality, longevity and catastrophe risk generated by our toolbox with the SCR under the Solvency II standard model.

stat.AP↗

arXiv2023

Bridging the gap between pricing and reserving with an occurrence and development model for non-life insurance claims

Due to the presence of reporting and settlement delay, claim data sets collected by non-life insurance companies are typically incomplete, facing right censored claim count and claim severity observations. Current practice in non-life insurance pricing tackles these right censored data via a two-step procedure. First, best estimates are computed for the number of claims that occurred in past exposure periods and the ultimate claim severities, using the incomplete, historical claim data. Second, pricing actuaries build predictive models to estimate technical, pure premiums for new contracts by treating these best estimates as actual observed outcomes, hereby neglecting their inherent uncertainty. We propose an alternative approach that brings valuable insights for both non-life pricing as well as reserving. As such we effectively bridge these two key actuarial tasks that have traditionally been discussed in silos. Hereto we develop a granular occurrence and development model for non-life claims that tackles reserving and at the same time resolves the inconsistency in traditional pricing techniques between actual observations and imputed best estimates. We illustrate our proposed model on an insurance as well as a reinsurance portfolio. The advantages of our proposed strategy are most compelling in the reinsurance illustration where large uncertainties in the best estimates originate from long reporting and settlement delays, low claim frequencies and heavy (even extreme) claim sizes.

q-fin.RM↗

arXiv2022

Insurance pricing with hierarchically structured data: An illustration with a workers' compensation insurance portfolio

Actuaries use predictive modeling techniques to assess the loss cost on a contract as a function of observable risk characteristics. State-of-the-art statistical and machine learning methods are not well equipped to handle hierarchically structured risk factors with a large number of levels. In this paper, we demonstrate the data-driven construction of an insurance pricing model when hierarchically structured risk factors, contract-specific as well as externally collected risk factors are available. We examine the pricing of a workers' compensation insurance product with a hierarchical credibility model (Jewell, 1975), Ohlsson's combination of a generalized linear and a hierarchical credibility model (Ohlsson, 2008) and mixed models. We compare the predictive performance of these models and evaluate the effect of the distributional assumption on the target variable by comparing linear mixed models with Tweedie generalized linear mixed models. For our case-study the Tweedie distribution is well suited to model and predict the loss cost on a contract. Moreover, incorporating contract-specific risk factors in the model improves the predictive performance and the risk differentiation in our workers' compensation insurance portfolio.

stat.AP↗

arXiv2021

A hierarchical reserving model for reported non-life insurance claims

Traditional non-life reserving models largely neglect the vast amount of information collected over the lifetime of a claim. This information includes covariates describing the policy, claim cause as well as the detailed history collected during a claim's development over time. We present the hierarchical reserving model as a modular framework for integrating a claim's history and claim-specific covariates into the development process. Hierarchical reserving models decompose the joint likelihood of the development process over time. Moreover, they are tailored to the portfolio at hand by adding a layer to the model for each of the events registered during the development of a claim (e.g. settlement, payment). Layers are modelled with statistical learning (e.g. generalized linear models) or machine learning methods (e.g. gradient boosting machines) and use claim-specific covariates. As a result of its flexibility, this framework incorporates many existing reserving models, ranging from aggregate models designed for run-off triangles to individual models using claim-specific covariates. This connection allows us to develop a data-driven strategy for choosing between aggregate and individual reserving; an important decision for reserving practitioners. We illustrate our method with a case study on a real insurance data set and deduce new insights in the covariates driving the development of claims. Moreover, we evaluate the method's performance on a large number of simulated portfolios representing several realistic development scenarios and demonstrate the flexibility and robustness of the hierarchical reserving model.

q-fin.RM↗

arXiv2021

Assessing the impact of the COVID-19 shock on a stochastic multi-population mortality model

We aim to assess the impact of a pandemic data point on the calibration of a stochastic multi-population mortality projection model and its resulting projections for future mortality rates. Throughout the paper we put focus on the Li & Lee mortality model, which has become a standard for projecting mortality in Belgium and the Netherlands. We calibrate this mortality model on annual deaths and exposures at the level of individual ages. This type of mortality data is typically collected, produced and reported with a significant delay of -- for some countries -- several years on a platform such as the Human Mortality Database. To enable a timely evaluation of the impact of a pandemic data point we have to rely on other data sources (e.g. the Short-Term Mortality Fluctuations Data series) that swiftly publish weekly mortality data collected in age buckets. To be compliant with the design and calibration strategy of the Li & Lee model, we have to transform the weekly mortality data collected in age buckets to yearly, age-specific observations. Therefore, our paper constructs a protocol to ungroup the deaths and exposures registered in age buckets to individual ages. To evaluate the impact of a pandemic shock, like COVID-19 in the year 2020, we weigh this data point in either the calibration or projection step. Obviously, the more weight we place on this data point, the more impact we observe on future estimated mortality rates and life expectancies. Our paper allows to quantify this impact and provides actuaries and actuarial associations with a framework to generate scenarios of future mortality under various assessments of the pandemic data point.

q-fin.RM↗

arXiv2021

Modeling the occurrence of events subject to a reporting delay via an EM algorithm

A delay between the occurrence and the reporting of events often has practical implications such as for the amount of capital to hold for insurance companies, or for taking preventive actions in case of infectious diseases. The accurate estimation of the number of incurred but not (yet) reported events forms an essential part of properly dealing with this phenomenon. We review the current practice for analysing such data and we present a flexible regression framework to jointly estimate the occurrence and reporting of events. By linking this setting to an incomplete data problem, estimation is performed via an expectation-maximization algorithm. The resulting method is elegant, easy to understand and implement, and provides refined insights in the nowcasts. The proposed methodology is applied to a European general liability portfolio in insurance.

stat.AP↗

arXiv2020

When stakes are high: balancing accuracy and transparency with Model-Agnostic Interpretable Data-driven suRRogates

Highly regulated industries, like banking and insurance, ask for transparent decision-making algorithms. At the same time, competitive markets are pushing for the use of complex black box models. We therefore present a procedure to develop a Model-Agnostic Interpretable Data-driven suRRogate (maidrr) suited for structured tabular data. Knowledge is extracted from a black box via partial dependence effects. These are used to perform smart feature engineering by grouping variable values. This results in a segmentation of the feature space with automatic variable selection. A transparent generalized linear model (GLM) is fit to the features in categorical format and their relevant interactions. We demonstrate our R package maidrr with a case study on general insurance claim frequency modeling for six publicly available datasets. Our maidrr GLM closely approximates a gradient boosting machine (GBM) black box and outperforms both a linear and tree surrogate as benchmarks.

stat.ML↗

arXiv2020

Sparse Regression with Multi-type Regularized Feature Modeling

Within the statistical and machine learning literature, regularization techniques are often used to construct sparse (predictive) models. Most regularization strategies only work for data where all predictors are treated identically, such as Lasso regression for (continuous) predictors treated as linear effects. However, many predictive problems involve different types of predictors and require a tailored regularization term. We propose a multi-type Lasso penalty that acts on the objective function as a sum of subpenalties, one for each type of predictor. As such, we allow for predictor selection and level fusion within a predictor in a data-driven way, simultaneous with the parameter estimation process. We develop a new estimation strategy for convex predictive models with this multi-type penalty. Using the theory of proximal operators, our estimation procedure is computationally efficient, partitioning the overall optimization problem into easier to solve subproblems, specific for each predictor type and its associated penalty. Earlier research applies approximations to non-differentiable penalties to solve the optimization problem. The proposed SMuRF algorithm removes the need for approximations and achieves a higher accuracy and computational efficiency. This is demonstrated with an extensive simulation study and the analysis of a case-study on insurance pricing analytics.

stat.CO↗

arXiv2020

Social network analytics for supervised fraud detection in insurance

Insurance fraud occurs when policyholders file claims that are exaggerated or based on intentional damages. This contribution develops a fraud detection strategy by extracting insightful information from the social network of a claim. First, we construct a network by linking claims with all their involved parties, including the policyholders, brokers, experts, and garages. Next, we establish fraud as a social phenomenon in the network and use the BiRank algorithm with a fraud specific query vector to compute a fraud score for each claim. From the network, we extract features related to the fraud scores as well as the claims' neighborhood structure. Finally, we combine these network features with the claim-specific features and build a supervised model with fraud in motor insurance as the target variable. Although we build a model for only motor insurance, the network includes claims from all available lines of business. Our results show that models with features derived from the network perform well when detecting fraud and even outperform the models using only the classical claim-specific features. Combining network and claim-specific features further improves the performance of supervised learning models to detect fraud. The resulting model flags highly suspicions claims that need to be further investigated. Our approach provides a guided and intelligent selection of claims and contributes to a more effective fraud investigation process.

cs.SI↗

arXiv2020

Pricing service maintenance contracts using predictive analytics

As more manufacturers shift their focus from selling products to end solutions, full-service maintenance contracts gain traction in the business world. These contracts cover all maintenance related costs during a predetermined horizon in exchange for a fixed service fee and relieve customers from uncertain maintenance costs. To guarantee profitability, the service fees should at least cover the expected costs during the contract horizon. As these expected costs may depend on several machine-dependent characteristics, e.g. operational environment, the service fees should also be differentiated based on these characteristics. If not, customers that are less prone to high maintenance costs will not buy into or renege on the contract. The latter can lead to adverse selection and leave the service provider with a maintenance-heavy portfolio, which may be detrimental to the profitability of the service contracts. We contribute to the literature with a data-driven tariff plan based on the calibration of predictive models that take into account the different machine profiles. This conveys to the service provider which machine profiles should be attracted at which price. We demonstrate the advantage of a differentiated tariff plan and show how it better protects against adverse selection.

stat.AP↗