Student Theses

Amato, Alessandra (2023) Applications of Early Warning Systems for Customer Segmentation of Wholesale Banking Clients. Abstract Wholesale Banking (WB) is evolving rapidly, and financial institutions are increasingly relying on Early Warning Systems (EWS) to detect financial distress proactively. For instance, ING has developed the Advanced Risk Integrated Application (ARIA) as its own EWS tool, primarily designed to detect ongoing negative changes. Nevertheless, ING aims at expanding ARIA’s capabilities and introducing the identification of potential up-selling opportunities within the artifact. To achieve this, this study investigates on the implementation of a novel Customer Segmentation (CS) model, integrating early warning triggers, using two unsupervised methods, namley K-Means and DBSCAN. The interpretation of the clusters generated is performed through a number of analyses that explore the different segments from multiple aspects, such as the tightness and separation of the subgroups formed, the clusters densities, the descriptive statistics of the customers’ distribution within each segment and risk-reward analyses. In conclusion, the research contributes to obtaining a deeper understanding of the financial health of banks’ WB clients, enabling them to deliver more targeted services based on each segment emerging needs. ...

Chen, D.T. (2023) Development of Financial Distress Prediction Model for the Watchlist Classification of Wholesale Banking Clients at ING. Abstract An Early Warning System (EWS) is a tool that enables the monitoring of the credit portfolio to identify clients in financial distress. ARIA is the EWS used by ING to monitor their Wholesale Banking (WB) clients using a variety of early warning triggers based on internal data, news articles, and market data. However, the current triggers are limited in their predictive capabilities as they are backwards-looking and are only derived from a single variable. A new Watchlist (WL) trigger aims to incorporate the information of all the current triggers into a single model that can predict whether a client should be on a watchlist based on their credit risk. The aim of this research focuses on exploring how such a WL trigger by developing a financial distress prediction model using machine learning techniques. ...

Kok, Dyon (2024) Stakeholder-centric approach to applying machine learning to probability of default models. Abstract This thesis explores the integration of Explainable Machine Learning (XML) into Probability of Default (PD) models, focusing on the Explainable Boosting Machine (EBM) and Generalized Additive Models with Interactions Network (GAMINet), compared against traditional logistic regression within a financial institution. It assesses explainability, regulatory compliance, performance, and operational feasibility, employing thorough model evaluation methods including architecture review, hyperparameter optimisation, and cross-validation. Despite not surpassing the existing logistic regression model in performance, the research underscores XML’s capacity to enhance PD models’ interpretability without sacrificing accuracy. It highlights EBM’s simplicity and GAMINet’s ability to capture complex interactions, with a particular emphasis on the significance of transparency and reasoned feature selection for stakeholder trust. The study also considers computational and integration challenges, ethical concerns, and bias management, advocating for a balanced XML integration strategy that combines EBM’s efficiency with GAMINet’s analytical depth. Future research directions include Python-SAS integration and advanced feature transformation to further refine model accuracy and interpretability. This work contributes to the broader conversation on leveraging machine learning in finance, striving to align technological progress with regulatory and ethical standards. ...

Beltman, J.F. (2023) Developing an Early Warning System for Retail Customer Deterioration: A Data-Driven Approach. Abstract This thesis addresses the existing gap in literature concerning the prevention of retail customers from becoming overdue. The literature gap in the area of retail EWS is the lack of a comprehensive approach that incorporates multiple data sources instead of relying on a single data source, along with the need for alternative methods, such as behavioral characteristics, for identifying deteriorating clients. The purpose of this thesis is to present the design and evaluation of an EWS for retail loans. Considering that financial institutions are often subject to regulations, this thesis avoids the use of algorithms that are considered poorly interpretable. The algorithms employed in this study include logistic regression, XGBoost, and Random Forest. Additionally, a fourth model, so called meta-model, is created by using the output of the other models, as input for a new model, which is trained by a logistic regression model. Regarding overall performance, the random forest algorithm achieves an average AUC of 0.775, while XGBoost and logistic regression achieve an AUC of 0.75. The meta-model outperforms the individual models with an AUC of 0.80. Compared to the random forest model, the XGBoost and logistic regression model did fit more on their top features. The meta-model consistently achieves the best fit when using the random forest predictions, as it is the top-performing individual algorithm. XGBoost serves as the second algorithm, while the logistic regression model provides the least significant contribution to the meta-model in terms of importance. The top-performing meta-model is utilized to gain practical insights into the timeliness of the EWS signals. The model reveals that a higher threshold for warning signals results in alerts closer to the overdue date, indicating increased sensitivity to emerging client deterioration. Conversely, lower thresholds focus more on the client’s overall status. Furthermore, using the top ten features for training yields satisfactory overall results, but incorporating features beyond the top ten provides valuable supplementary information. ...

REil, Jens (2025). Evaluating the Impact of Integration of ESG Factors in Credit Ratings Using External Credit Assessment Institutions for Corporates Abstract This thesis examines the integration of Environmental, Social, and Governance (ESG) factors into credit ratings, focusing on the methodologies employed by External Credit As sessment Institutions (ECAIs) and their implications for financial risk models, particularly in the banking sector. As ESG considerations gain increasing importance in investment decisions, understanding how these factors are incorporated into credit ratings has become critical for financial institutions like ING. This study investigates the varying approaches to ESG integration across different ECAIs and evaluates the challenges and opportunities posed by their respective methodologies. By conducting a literature review and assess ing current practices, the thesis identifies ESG risks through ECAI methodologies and the important variables driving ESG risks to be considered within models. The research concludes by proposing actionable heuristics for financial institutions to better align their credit risk models with ESG factors, aiming to enhance transparency, regulatory compli ance, and risk management strategies. The findings contribute to ESG integration in credit risk assessments, providing actionable insights for financial institutions like ING to adapt to evolving market expectations and regulatory demands ...

Goldmann, Sebastian Hendrikus (2024). Enhancing Credit Risk Prediction in Retail Banking: Integrating Time Series and Classical ML Algorithms. Abstract This thesis investigates the application of Time Series Classification (TSC) algorithms to enhance credit risk models, focusing on the historical balance data of retail customers. The primary aim was to explore how TSC models could improve the discriminatory power of these risk models. Employing various TSC techniques, including deep learning, shapelets, and Canonical Interval Forest (CIF) algorithms, the study rigorously evaluated their performance in predicting the Probability of Default (PD). Additionally, this thesis contributes to the evolving field of credit risk modeling by introducing a novel approach that integrates TSC with traditional credit risk assessment methods. The research revealed that TSC algorithms, particularly when applied to end-of-day balance data, have the potential to significantly enhance the predictive accuracy of credit risk models. The CIF model, in particular, demonstrated notable efficacy, rivaling the performance of existing credit risk models. However, applying TSC algorithms to multivariate monthly data showed limited effectiveness, suggesting the removal of critical information in such data aggregation. The study also highlighted the interpretability challenges with complex TSC models and the need for more holistic data inclusion for a comprehensive credit risk assessment. ...

Pol, Stijn van der (2024). The Creation of an Explainable Artificial Intelligence Model to Enhance Interpretability and Transparency for ING in Their Fight Against Fraud Abstract This thesis presents the development of an Explainable Artificial Intelligence (XAI) model aimed at enhancing the interpretability and transparency of fraud detection systems at ING. By integrating XAI techniques, the model provides clear and actionable insights into the decision-making processes of machine learning classifiers, thereby bridging the gap between high model accuracy and the need for understandable outputs. Through extensive experimentation and validation with real-world data, this research demonstrates the practical applicability of XAI in transactional fraud detection. The insights gained from expert interviews and use case testing further reinforce the model’s effectiveness in improving stakeholder communication and decision-making. This work contributes to the ongoing efforts to develop AI systems that are not only powerful but also transparent and trustworthy in combating financial fraud. ...

Koene, Thomas (2025). Navigating Human Rights Due Diligence in Financial Institutions: Creating A Roadmap for Implementation Abstract This research investigates how financial institutions (FIs) can effectively implement Human Rights Due Diligence (HRDD) on their direct business partners. Both the downstream and up stream part of the value chain is covered. A systematic literature review (SLR) is used to com bine the existing academic and gray literature on the HRDD process, the CSDDD and its key parts, current guidelines and frameworks, and the Action Research Model by Warner Burke. ComplementingtheSLR,theresearch utilizes a mixed-method approach featuring semi struc tured interviews (SSIs) with industry experts from a leading financial institution, from now on Company X. Empirical data reveals challenges, opportunities and best-practices for both im plementing HRDDandperformingHRDDitself. Theresults show the necessity of clear defini tions, stakeholder engagement, standardization of processes, and integration with existing risk and compliance systems. Drawing on the Action Research Model by Warner Burke, the study develops a roadmap that provides a high-level, step-by-step guide to HRDD implementation ondirect business partners, thereby bridging theory and practice. ...

Deep Learning for Financial Time Series Forecasting Student: Vitalii Fishchuk Program: MSc Business Information Technology, University of Twente Period: 2023-2024 Abstract This thesis investigates the application of deep learning techniques for financial time series forecasting at ING Bank. The research explores various neural network architectures including Long Short-Term Memory (LSTM) networks, Transformer models, and hybrid approaches for predicting financial metrics. The study evaluates model performance across different forecasting horizons and compares deep learning methods with traditional statistical approaches. Special attention is given to the interpretability of model predictions in the context of banking operations and regulatory requirements. ...