From Statistical Interpretations to Explainable AI in Machine Learning - Enhancing Decision-Making in the Lending Industry

Aug 3, 2026

Upload by Maeve

The transparency and interpretability of modelling are crucial in high-risk applications. In lending, regulators worldwide require interpretable decision-making models, sustaining the practical role of inherently interpretable statistical approaches. Meanwhile, black-box machine learning offers strong predictive accuracy, driving research into interpretability. This thesis investigates decision-making in the lending industry using traditional statistical models and post-hoc explanation methods, focusing on LIME and SHAP to characterize model behaviour through input attributions and output relationships.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages201

Reading Time8 h 27 min

AccessFree to download

Abstract

Interpretability requirements in lending

Post-hoc interpretation methods (LIME, SHAP)

Focus of the thesis: statistical models and explainable AI

Paper 1: probit models and COVID-19 firm financing

Paper 2: interpretable approaches for credit scoring stability

This thesis has been submitted in fulfilment of the requirements for a postgraduate degree (e. g. PhD, MPhil, DClinPsychol) at the University of Edinburgh. Please note the following terms and conditions of use: • This work is protected by copyright and other intellectual property rights, which are retained by the thesis author, unless otherwise stated. • A copy can be downloaded for personal non-commercial research or study, without prior permission or charge. • This thesis cannot be reproduced or quoted extensively from without first obtaining permission in writing from the author. • The content must not be changed in any way or sold commercially in any format or medium without the formal permission of the author. • When referring to this work, full bibliographic details including the author, title, awarding institution and date of the thesis must be given. From Statistical Interpretations to Explainable AI in Machine Learning: Enhancing Decision-Making in the Lending Industry Yujia Chen Doctor of Philosophy The University of Edinburgh 2024 Declaration This thesis has been composed by myself and contains no material that has been accepted for the award of any other degree at any university. Parts of this thesis have been published in the following journals: • Cowling, M., Liu, W., Chen, Y., Calabrese, R., and Vorley, T. (2023) . Financing small and innovative firms during COVID-19 . Economics of Innovation and New Technology, 1-28 . • Chen, Y., Calabrese, R., and Martin-Barragan, B. (2024) . Interpretable machine learning for imbalanced credit scoring datasets. European Journal of Operational Research, 312 (1), 357–372 . Permission to include text from these papers has been gained from the publisher and the authors. To the best of my knowledge and belief this thesis contains no other material previously published by any other person except where due acknowledgement has been made. (Yujia Chen) Acknowledgements When I started my PhD, I thought it would be a lonely and arduous journey. Looking back now, I realise that it is filled with challenges, but I am truly fortunate to have a group of wonderful people supporting and accompanying me throughout. I am profoundly grateful for their presence. I would like to express my gratitude to my supervisor, Belen Martin-Barragan, who solidifies my determination to pursue research. Many thanks for her generosity in sharing her experiences and knowledge with me. The patience, attention, and guidance she offered during our discussions in her office will remain etched in my memory. I deeply appreciate her encouragement and her belief in me, which have been instrumental in recognising my strengths. My sincere appreciation also goes to my co-supervisor, Raffaella Calabrese, for her expert guidance and invaluable support in my doctoral journey, and for her always having faith in me, which has motivated me to move forward bravely. Her professionalism, enthusiasm, and energy have deeply inspired me. I am extremely grateful to have had the opportunity to work with her and have learned so much from her. I would like to thank all my colleagues and co-authors for their stimulating discussions and collaboration. My appreciation also extends to the University of Edinburgh for providing the necessary resources and a conducive research environment. My heartfelt gratitude also goes to my friends, who have shared in my joys and excitement as well as my anxieties and stresses. Their presence has made this journey much more enjoyable. Last but not least, I am deeply indebted to my parents, for their immense love and unwavering respect, support, and understanding throughout this journey. I feel so lucky to be their daughter. Abstract The transparency and interpretability of modelling are crucial in high-risk applications. For example, financial regulators worldwide have mandated the necessity for interpretability in decision-making models within the lending s

FAQs

Why is interpretability important in lending decision-making?

Lending is a high-risk domain where regulators worldwide mandate interpretability. This requirement supports the continued use of inherently interpretable modelling approaches.

How do LIME and SHAP help explain black-box models?

LIME and SHAP generate post-hoc explanations that reveal model behaviour after predictions. They do so by analyzing relationships between input attributions and model outputs.

What does the thesis examine in the context of the lending industry?

The thesis studies decision-making using both traditional statistical models and post-hoc interpretation methods. It applies probit models and interpretability-focused approaches to analyze financing outcomes.

Aug 3, 2026

Upload by Maeve