Tim Verdonck

Jakob Raymaekers, Peter J. Rousseeuw, Tim Verdonck et al.

Linear model trees are regression trees that incorporate linear models in the leaf nodes. This preserves the intuitive interpretation of decision trees and at the same time enables them to better capture linear relationships, which is hard for standard decision trees. But most existing methods for fitting linear model trees are time consuming and therefore not scalable to large data sets. In addition, they are more prone to overfitting and extrapolation issues than standard regression trees. In this paper we introduce PILOT, a new algorithm for linear model trees that is fast, regularized, stable and interpretable. PILOT trains in a greedy fashion like classic regression trees, but incorporates an $L^2$ boosting approach and a model selection rule for fitting linear models in the nodes. The abbreviation PILOT stands for $PI$ecewise $L$inear $O$rganic $T$ree, where `organic' refers to the fact that no pruning is carried out. PILOT has the same low time and space complexity as CART without its pruning. An empirical study indicates that PILOT tends to outperform standard decision trees and other linear model trees on a variety of data sets. Moreover, we prove its consistency in an additive model setting under weak assumptions. When the data is generated by a linear model, the convergence rate is polynomial.

Toon Vanderschueren, Robert Boute, Tim Verdonck et al.

Machine maintenance is a challenging operational problem, where the goal is to plan sufficient preventive maintenance to avoid machine failures and overhauls. Maintenance is often imperfect in reality and does not make the asset as good as new. Although a variety of imperfect maintenance policies have been proposed in the literature, these rely on strong assumptions regarding the effect of maintenance on the machine's condition, assuming the effect is (1) deterministic or governed by a known probability distribution, and (2) machine-independent. This work proposes to relax both assumptions by learning the effect of maintenance conditional on a machine's characteristics from observational data on similar machines using existing methodologies for causal inference. By predicting the maintenance effect, we can estimate the number of overhauls and failures for different levels of maintenance and, consequently, optimize the preventive maintenance frequency to minimize the total estimated cost. We validate our proposed approach using real-life data on more than 4,000 maintenance contracts from an industrial partner. Empirical results show that our novel, causal approach accurately predicts the maintenance effect and results in individualized maintenance schedules that are more accurate and cost-effective than supervised or non-individualized approaches.

Christopher Bockel-Rickermann, Tim Verdonck, Wouter Verbeke

The literature on fraud analytics and fraud detection has seen a substantial increase in output in the past decade. This has led to a wide range of research topics and overall little organization of the many aspects of fraud analytical research. The focus of academics ranges from identifying fraudulent credit card payments to spotting illegitimate insurance claims. In addition, there is a wide range of methods and research objectives. This paper aims to provide an overview of fraud analytics in research and aims to more narrowly organize the discipline and its many subfields. We analyze a sample of almost 300 records on fraud analytics published between 2011 and 2020. In a systematic way, we identify the most prominent domains of application, challenges faced, performance metrics, and methods used. In addition, we build a framework for fraud analytical methods and propose a keywording strategy for future research. One of the key challenges in fraud analytics is access to public datasets. To further aid the community, we provide eight requirements for suitable data sets in research motivated by our research. We structure our sample of the literature in an online database. The database is available online for fellow researchers to investigate and potentially build upon.

Yves-Cédric Bauwelinckx, Jan Dhaene, Tim Verdonck et al.

Modeling lies at the core of both the financial and the insurance industry for a wide variety of tasks. The rise and development of machine learning and deep learning models have created many opportunities to improve our modeling toolbox. Breakthroughs in these fields often come with the requirement of large amounts of data. Such large datasets are often not publicly available in finance and insurance, mainly due to privacy and ethics concerns. This lack of data is currently one of the main hurdles in developing better models. One possible option to alleviating this issue is generative modeling. Generative models are capable of simulating fake but realistic-looking data, also referred to as synthetic data, that can be shared more freely. Generative Adversarial Networks (GANs) is such a model that increases our capacity to fit very high-dimensional distributions of data. While research on GANs is an active topic in fields like computer vision, they have found limited adoption within the human sciences, like economics and insurance. Reason for this is that in these fields, most questions are inherently about identification of causal effects, while to this day neural networks, which are at the center of the GAN framework, focus mostly on high-dimensional correlations. In this paper we study the causal preservation capabilities of GANs and whether the produced synthetic data can reliably be used to answer causal questions. This is done by performing causal analyses on the synthetic data, produced by a GAN, with increasingly more lenient assumptions. We consider the cross-sectional case, the time series case and the case with a complete structural model. It is shown that in the simple cross-sectional scenario where correlation equals causation the GAN preserves causality, but that challenges arise for more advanced analyses.

Christopher Bockel-Rickermann, Toon Vanderschueren, Jeroen Berrevoets et al.

Estimating a unit's responses to interventions with an associated dose, the "conditional average dose response" (CADR), is relevant in a variety of domains, from healthcare to business, economics, and beyond. Such a response typically needs to be estimated from observational data, which introduces several challenges. That is why the machine learning (ML) community has proposed several tailored CADR estimators. Yet, the proposal of most of these methods requires strong assumptions on the distribution of data and the assignment of interventions, which go beyond the standard assumptions in causal inference. Whereas previous works have so far focused on smooth shifts in covariate distributions across doses, in this work, we will study estimating CADR from clustered data and where different doses are assigned to different segments of a population. On a novel benchmarking dataset, we show the impacts of clustered data on model performance and propose an estimator, CBRNet, that learns cluster-agnostic and hence dose-agnostic covariate representations through representation balancing for unbiased CADR inference. We run extensive experiments to illustrate the workings of our method and compare it with the state of the art in ML for CADR estimation.

Bruno Deprez, Wouter Verbeke, Tim Verdonck

Modelling claim frequency and severity for non-life insurance pricing predominantly relies on generalised linear models, with gradient-boosted machines as the leading machine learning alternative. Tabular foundation models (TFMs) offer a fundamentally different paradigm. By pre-training on large collections of synthetic datasets, TFMs enable inference on new data through in-context learning, without any dataset-specific fitting or hyperparameter tuning. This paper presents a first empirical evaluation of TabPFN for motor insurance pricing, benchmarking it against GLM and XGBoost on two publicly available MTPL datasets. Our results show that TabPFN does not consistently outperform established baselines, exhibits substantially longer inference times, and is sensitive to the size of the in-context training set. While tabular foundation models represent a promising direction, particularly in data-scarce settings, their current formulation does not offer a viable replacement for established actuarial methods.

Christopher Bockel-Rickermann, Sam Verboven, Tim Verdonck et al.

In lending, where prices are specific to both customers and products, having a well-functioning personalized pricing policy in place is essential to effective business making. Typically, such a policy must be derived from observational data, which introduces several challenges. While the problem of ``endogeneity'' is prominently studied in the established pricing literature, the problem of selection bias (or, more precisely, bid selection bias) is not. We take a step towards understanding the effects of selection bias by posing pricing as a problem of causal inference. Specifically, we consider the reaction of a customer to price a treatment effect. In our experiments, we simulate varying levels of selection bias on a semi-synthetic dataset on mortgage loan applications in Belgium. We investigate the potential of parametric and nonparametric methods for the identification of individual bid-response functions. Our results illustrate how conventional methods such as logistic regression and neural networks suffer adversely from selection bias. In contrast, we implement state-of-the-art methods from causal machine learning and show their capability to overcome selection bias in pricing data.

Bart Baesens, Andreas Goethals, Stefan Lessmann et al.

Predictive models play a pivotal role in credit risk management, guiding critical decisions through accurate estimation of default probabilities and losses. Extensive research has introduced new modeling techniques, complemented by large-scale benchmarking studies consolidating the state-of-the-art. Today, quasi-standards such as gradient-boosting models paired with SHAP explainers have emerged, yet continuous improvement of risk models remains a top priority. Concurrently, rapid advancements in AI, most notably large language models, have disrupted predictive modeling paradigms. Foundation models, pretrained on extensive datasets from diverse domains, have demonstrated remarkable performance by leveraging prior knowledge. While prevalent in natural language processing and computer vision, foundation models for tabular data have only recently emerged. We conjecture that pretraining on out-of-domain data is particularly beneficial in small-data settings, such as SME lending or specialized corporate portfolios, and may help address longstanding challenges including low default portfolios and class imbalance. This paper benchmarks recently proposed tabular foundation models against a broad set of competitors, including established and advanced machine learning techniques, across two core tasks: PD and LGD modeling. Our evaluation encompasses various datasets, performance indicators, and experimental conditions. We find that tabular foundation models generally perform best across datasets and tasks. Moreover, they offer significant improvement in predictive performance as dataset size shrinks. These results are remarkable given that the models are tested out-of-the-box, without hyperparameter tuning, ensuring ease of use and mitigating computational costs.

Félix Vandervorst, Bruno Deprez, Wouter Verbeke et al.

Graph-based methods are becoming increasingly popular in machine learning due to their ability to model complex data and relations. Insurance fraud is a prime use case, since false claims are often the result of organised criminals that stage accidents or the same persons filing erroneous claims on multiple policies. One challenge is that graph-based approaches struggle to find meaningful representations of the data because of the high class imbalance present in fraud data. Another is that insurance networks are heterogeneous and dynamic, given the changing relations among people, companies and policies. That is why gradient boosted tree approaches on tabular data still dominate the field. Therefore, we present a novel inductive graph gradient boosting machine (G-GBM) for supervised learning on heterogeneous and dynamic graphs. We show that our estimator competes with popular graph neural network approaches in an experiment using a variety of simulated random graphs. We demonstrate the power of G-GBM for insurance fraud detection using an open-source and a real-world, proprietary dataset. Given that the backbone model is a gradient boosting forest, we apply established explainability methods to gain better insights into the predictions made by G-GBM.

Bruno Deprez, Bart Baesens, Tim Verdonck et al.

Purpose: This paper introduces a novel graph-based method, GARG-AML, for efficient and effective anti-money laundering (AML). It quantifies smurfing risk, a popular money laundering method, by providing each node in the network with a single interpretable score. The proposed method strikes a balance among computational efficiency, detection power and transparency. Different versions of GARG-AML are introduced for undirected and directed networks. Methodology: GARG-AML constructs the adjacency matrix of a node's second-order neighbourhood in a specific way. This allows us to use the density of different blocks in the adjacency matrix to express the neighbourhood's resemblance to a pure smurfing pattern. GARG-AML is extended using a decision tree and gradient-boosting classifier to increase its performance even more. The methods are tested on synthetic and on open-source data against the current state-of-the-art in AML. Findings: We find that GARG-AML obtains state-of-the-art performance on all datasets. We illustrate that GARG-AML scales well to massive transactions graphs encountered at financial institutions. By leveraging only the adjacency matrix of the second-order neighbourhood and basic network features, this work highlights the potential of fundamental network properties towards advancing fraud detection. Originality: This paper uses only basic network features and expert knowledge on smurfing to construct a performant AML system. The originality lies in the translation of smurfing detection to these features and network representation. Our proposed method is built around the real business needs of scalability and interpretability. It therefore provides a solution that can be easily implemented at financial institutions or incorporated in existing AML solutions.

Nick Berlanger, Noah van Ophoven, Tim Verdonck et al.

Accurate forecasts for day-ahead photovoltaic (PV) power generation are crucial to support a high PV penetration rate in the local electricity grid and to assure stability in the grid. We use state-of-the-art tree-based machine learning methods to produce such forecasts and, unlike previous studies, we hereby account for (i) the effects various meteorological as well as astronomical features have on PV power production, and this (ii) at coarse as well as granular spatial locations. To this end, we use data from Belgium and forecast day-ahead PV power production at an hourly resolution. The insights from our study can assist utilities, decision-makers, and other stakeholders in optimizing grid operations, economic dispatch, and in facilitating the integration of distributed PV power into the electricity grid.

Steven Mortier, Amir Hamedpour, Bart Bussmann et al.

Changes in climate can greatly affect the phenology of plants, which can have important feedback effects, such as altering the carbon cycle. These phenological feedback effects are often induced by a shift in the start or end dates of the growing season of plants. The normalized difference vegetation index (NDVI) serves as a straightforward indicator for assessing the presence of green vegetation and can also provide an estimation of the plants' growing season. In this study, we investigated the effect of soil temperature on the timing of the start of the season (SOS), timing of the peak of the season (POS), and the maximum annual NDVI value (PEAK) in subarctic grassland ecosystems between 2014 and 2019. We also explored the impact of other meteorological variables, including air temperature, precipitation, and irradiance, on the inter-annual variation in vegetation phenology. Using machine learning (ML) techniques and SHapley Additive exPlanations (SHAP) values, we analyzed the relative importance and contribution of each variable to the phenological predictions. Our results reveal a significant relationship between soil temperature and SOS and POS, indicating that higher soil temperatures lead to an earlier start and peak of the growing season. However, the Peak NDVI values showed just a slight increase with higher soil temperatures. The analysis of other meteorological variables demonstrated their impacts on the inter-annual variation of the vegetation phenology. Ultimately, this study contributes to our knowledge of the relationships between soil temperature, meteorological variables, and vegetation phenology, providing valuable insights for predicting vegetation phenology characteristics and managing subarctic grasslands in the face of climate change. Additionally, this work provides a solid foundation for future ML-based vegetation phenology studies.

Sarah Leyder, Jakob Raymaekers, Peter J. Rousseeuw et al.

Independent component analysis (ICA) is a powerful tool for decomposing a multivariate signal or distribution into fully independent sources, not just uncorrelated ones. Unfortunately, most approaches to ICA are not robust against outliers. Here we propose a robust ICA method called RICA, which estimates the components by minimizing a robust measure of dependence between multivariate random variables. The dependence measure used is the distance correlation (dCor). In order to make it more robust we first apply a new transformation called the bowl transform, which is bounded, one-to-one, continuous, and maps far outliers to points close to the origin. This preserves the crucial property that a zero dCor implies independence. RICA estimates the independent sources sequentially, by looking for the component that has the smallest dCor with the remainder. RICA is strongly consistent and has the usual parametric rate of convergence. Its robustness is investigated by a simulation study, in which it generally outperforms its competitors. The method is illustrated on three applications, including the well-known cocktail party problem.

Bruno Deprez, Wei Wei, Wouter Verbeke et al.

Financial institutions are required by regulation to report suspicious financial transactions related to money laundering. Therefore, they need to constantly monitor vast amounts of incoming and outgoing transactions. A particular challenge in detecting money laundering is that money launderers continuously adapt their tactics to evade detection. Hence, detection methods need constant fine-tuning. Traditional machine learning models suffer from catastrophic forgetting when fine-tuning the model on new data, thereby limiting their effectiveness in dynamic environments. Continual learning methods may address this issue and enhance current anti-money laundering (AML) practices, by allowing models to incorporate new information while retaining prior knowledge. Research on continual graph learning for AML, however, is still scarce. In this review, we critically evaluate state-of-the-art continual graph learning approaches for AML applications. We categorise methods into replay-based, regularization-based, and architecture-based strategies within the graph neural network (GNN) framework, and we provide in-depth experimental evaluations on both synthetic and real-world AML data sets that showcase the effect of the different hyperparameters. Our analysis demonstrates that continual learning improves model adaptability and robustness in the face of extreme class imbalances and evolving fraud patterns. Finally, we outline key challenges and propose directions for future research.

Jakob Raymaekers, Peter J. Rousseeuw, Thomas Servotte et al.

Random forests are widely used in regression. However, the decision trees used as base learners are poor approximators of linear relationships. To address this limitation we propose RaFFLE (Random Forest Featuring Linear Extensions), a novel framework that integrates the recently developed PILOT trees (Piecewise Linear Organic Trees) as base learners within a random forest ensemble. PILOT trees combine the computational efficiency of traditional decision trees with the flexibility of linear model trees. To ensure sufficient diversity of the individual trees, we introduce an adjustable regularization parameter and use node-level feature sampling. These modifications improve the accuracy of the forest. We establish theoretical guarantees for the consistency of RaFFLE under weak conditions, and its faster convergence when the data are generated by a linear model. Empirical evaluations on 136 regression datasets demonstrate that RaFFLE outperforms the classical CART and random forest methods, the regularized linear methods Lasso and Ridge, and the state-of-the-art XGBoost algorithm, across both linear and nonlinear datasets. By balancing predictive accuracy and computational efficiency, RaFFLE proves to be a versatile tool for tackling a wide variety of regression problems.

Christopher Bockel-Rickermann, Toon Vanderschueren, Tim Verdonck et al.

Estimating conditional average dose responses (CADR) is an important but challenging problem. Estimators must correctly model the potentially complex relationships between covariates, interventions, doses, and outcomes. In recent years, the machine learning community has shown great interest in developing tailored CADR estimators that target specific challenges. Their performance is typically evaluated against other methods on (semi-) synthetic benchmark datasets. Our paper analyses this practice and shows that using popular benchmark datasets without further analysis is insufficient to judge model performance. Established benchmarks entail multiple challenges, whose impacts must be disentangled. Therefore, we propose a novel decomposition scheme that allows the evaluation of the impact of five distinct components contributing to CADR estimator performance. We apply this scheme to eight popular CADR estimators on four widely-used benchmark datasets, running nearly 1,500 individual experiments. Our results reveal that most established benchmarks are challenging for reasons different from their creators' claims. Notably, confounding, the key challenge tackled by most estimators, is not an issue in any of the considered datasets. We discuss the major implications of our findings and present directions for future research.

Toon Vanderschueren, Bart Baesens, Tim Verdonck et al.

A central problem in business concerns the optimal allocation of limited resources to a set of available tasks, where the payoff of these tasks is inherently uncertain. In credit card fraud detection, for instance, a bank can only assign a small subset of transactions to their fraud investigations team. Typically, such problems are solved using a classification framework, where the focus is on predicting task outcomes given a set of characteristics. Resources are then allocated to the tasks that are predicted to be the most likely to succeed. However, we argue that using classification to address task uncertainty is inherently suboptimal as it does not take into account the available capacity. Therefore, we first frame the problem as a type of assignment problem. Then, we present a novel solution using learning to rank by directly optimizing the assignment's expected profit given limited, stochastic capacity. This is achieved by optimizing a specific instance of the net discounted cumulative gain, a commonly used class of metrics in learning to rank. Empirically, we demonstrate that our new method achieves higher expected profit and expected precision compared to a classification approach for a wide variety of application areas and data sets. This illustrates the benefit of an integrated approach and of explicitly considering the available resources when learning a predictive model.

Robin Van Oirbeek, Jolien Ponnet, Tim Verdonck

Performance measurement is an essential task once a statistical model is created. The Area Under the receiving operating characteristics Curve (AUC) is the most popular measure for evaluating the quality of a binary classifier. In this case, AUC is equal to the concordance probability, a frequently used measure to evaluate the discriminatory power of the model. Contrary to AUC, the concordance probability can also be extended to the situation with a continuous response variable. Due to the staggering size of data sets nowadays, determining this discriminatory measure requires a tremendous amount of costly computations and is hence immensely time consuming, certainly in case of a continuous response variable. Therefore, we propose two estimation methods that calculate the concordance probability in a fast and accurate way and that can be applied to both the discrete and continuous setting. Extensive simulation studies show the excellent performance and fast computing times of both estimators. Finally, experiments on two real-life data sets confirm the conclusions of the artificial simulations.

Jakob Raymaekers, Wouter Verbeke, Tim Verdonck

In many practical applications, such as fraud detection, credit risk modeling or medical decision making, classification models for assigning instances to a predefined set of classes are required to be both precise as well as interpretable. Linear modeling methods such as logistic regression are often adopted, since they offer an acceptable balance between precision and interpretability. Linear methods, however, are not well equipped to handle categorical predictors with high-cardinality or to exploit non-linear relations in the data. As a solution, data preprocessing methods such as weight-of-evidence are typically used for transforming the predictors. The binning procedure that underlies the weight-of-evidence approach, however, has been little researched and typically relies on ad-hoc or expert driven procedures. The objective in this paper, therefore, is to propose a formalized, data-driven and powerful method. To this end, we explore the discretization of continuous variables through the binning of spline functions, which allows for capturing non-linear effects in the predictor variables and yields highly interpretable predictors taking only a small number of discrete values. Moreover, we extend upon the weight-of-evidence approach and propose to estimate the proportions using shrinkage estimators. Together, this offers an improved ability to exploit both non-linear and categorical predictors for achieving increased classification precision, while maintaining interpretability of the resulting model and decreasing the risk of overfitting. We present the results of a series of experiments in a fraud detection setting, which illustrate the effectiveness of the presented approach. We facilitate reproduction of the presented results and adoption of the proposed approaches by providing both the dataset and the code for implementing the experiments and the presented approach.

Bart Baesens, Sebastiaan Höppner, Irene Ortner et al.

A major challenge when trying to detect fraud is that the fraudulent activities form a minority class which make up a very small proportion of the data set. In most data sets, fraud occurs in typically less than 0.5% of the cases. Detecting fraud in such a highly imbalanced data set typically leads to predictions that favor the majority group, causing fraud to remain undetected. We discuss some popular oversampling techniques that solve the problem of imbalanced data by creating synthetic samples that mimic the minority class. A frequent problem when analyzing real data is the presence of anomalies or outliers. When such atypical observations are present in the data, most oversampling techniques are prone to create synthetic samples that distort the detection algorithm and spoil the resulting analysis. A useful tool for anomaly detection is robust statistics, which aims to find the outliers by first fitting the majority of the data and then flagging data observations that deviate from it. In this paper, we present a robust version of ROSE, called robROSE, which combines several promising approaches to cope simultaneously with the problem of imbalanced data and the presence of outliers. The proposed method achieves to enhance the presence of the fraud cases while ignoring anomalies. The good performance of our new sampling technique is illustrated on simulated and real data sets and it is shown that robROSE can provide better insight in the structure of the data. The source code of the robROSE algorithm is made freely available.

Robin Van Oirbeek, Christopher Grumiau, Tim Verdonck

The concordance probability or C-index is a popular measure to capture the discriminatory ability of a regression model. In this article, the definition of this measure is adapted to the specific needs of the frequency and severity model, typically used during the technical pricing of a non-life insurance product. Due to the typical large sample size of the frequency data in particular, two different adaptations of the estimation procedure of the concordance probability are presented. Note that the latter procedures can be applied to all different versions of the concordance probability.

Sebastiaan Höppner, Eugen Stripling, Bart Baesens et al.

Customer retention campaigns increasingly rely on predictive models to detect potential churners in a vast customer base. From the perspective of machine learning, the task of predicting customer churn can be presented as a binary classification problem. Using data on historic behavior, classification algorithms are built with the purpose of accurately predicting the probability of a customer defecting. The predictive churn models are then commonly selected based on accuracy related performance measures such as the area under the ROC curve (AUC). However, these models are often not well aligned with the core business requirement of profit maximization, in the sense that, the models fail to take into account not only misclassification costs, but also the benefits originating from a correct classification. Therefore, the aim is to construct churn prediction models that are profitable and preferably interpretable too. The recently developed expected maximum profit measure for customer churn (EMPC) has been proposed in order to select the most profitable churn model. We present a new classifier that integrates the EMPC metric directly into the model construction. Our technique, called ProfTree, uses an evolutionary algorithm for learning profit driven decision trees. In a benchmark study with real-life data sets from various telecommunication service providers, we show that ProfTree achieves significant profit improvements compared to classic accuracy driven tree-based methods.