Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IQNN-CS: Interpretable Quantum Neural Network for Credit Scoring

Abdul Samad Khan, Nouhaila Innan, Aeysha Khalique, Muhammad Shafique

TL;DR

This work tackles interpretability in quantum neural networks for high-stakes credit scoring by proposing IQNN-CS, a hybrid quantum–classical pipeline that uses a variational QNN and post-hoc explanations. A key contribution is Inter-Class Attribution Alignment (ICAA), which quantifies how distinctly the model differentiates between risk classes by comparing class-attribution vectors in the input space. Evaluation on two real-world datasets shows strong performance when data align with the quantum encoding and reveals how attribution patterns and latent structure govern interpretability, with ICAA serving as a diagnostic tool for breakdowns in multiclass reasoning. The approach provides a practical path toward transparent, regulator-friendly QML in financial decision-making, illustrating both the potential and limitations of interpretable quantum models in structured, multiclass tasks.

Abstract

Credit scoring is a high-stakes task in financial services, where model decisions directly impact individuals' access to credit and are subject to strict regulatory scrutiny. While Quantum Machine Learning (QML) offers new computational capabilities, its black-box nature poses challenges for adoption in domains that demand transparency and trust. In this work, we present IQNN-CS, an interpretable quantum neural network framework designed for multiclass credit risk classification. The architecture combines a variational QNN with a suite of post-hoc explanation techniques tailored for structured data. To address the lack of structured interpretability in QML, we introduce Inter-Class Attribution Alignment (ICAA), a novel metric that quantifies attribution divergence across predicted classes, revealing how the model distinguishes between credit risk categories. Evaluated on two real-world credit datasets, IQNN-CS demonstrates stable training dynamics, competitive predictive performance, and enhanced interpretability. Our results highlight a practical path toward transparent and accountable QML models for financial decision-making.

IQNN-CS: Interpretable Quantum Neural Network for Credit Scoring

TL;DR

This work tackles interpretability in quantum neural networks for high-stakes credit scoring by proposing IQNN-CS, a hybrid quantum–classical pipeline that uses a variational QNN and post-hoc explanations. A key contribution is Inter-Class Attribution Alignment (ICAA), which quantifies how distinctly the model differentiates between risk classes by comparing class-attribution vectors in the input space. Evaluation on two real-world datasets shows strong performance when data align with the quantum encoding and reveals how attribution patterns and latent structure govern interpretability, with ICAA serving as a diagnostic tool for breakdowns in multiclass reasoning. The approach provides a practical path toward transparent, regulator-friendly QML in financial decision-making, illustrating both the potential and limitations of interpretable quantum models in structured, multiclass tasks.

Abstract

Credit scoring is a high-stakes task in financial services, where model decisions directly impact individuals' access to credit and are subject to strict regulatory scrutiny. While Quantum Machine Learning (QML) offers new computational capabilities, its black-box nature poses challenges for adoption in domains that demand transparency and trust. In this work, we present IQNN-CS, an interpretable quantum neural network framework designed for multiclass credit risk classification. The architecture combines a variational QNN with a suite of post-hoc explanation techniques tailored for structured data. To address the lack of structured interpretability in QML, we introduce Inter-Class Attribution Alignment (ICAA), a novel metric that quantifies attribution divergence across predicted classes, revealing how the model distinguishes between credit risk categories. Evaluated on two real-world credit datasets, IQNN-CS demonstrates stable training dynamics, competitive predictive performance, and enhanced interpretability. Our results highlight a practical path toward transparent and accountable QML models for financial decision-making.
Paper Structure (24 sections, 8 figures, 5 tables, 2 algorithms)

This paper contains 24 sections, 8 figures, 5 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. QML for Credit Scoring
  4. Interpretability in QML
  5. Methodology
  6. Data Preparation
  7. Hybrid Quantum-Classical Model
  8. Training Procedure
  9. Post-Hoc Interpretability
  10. Evaluation Metrics
  11. Results
  12. Experimental Settings
  13. Training and Convergence
  14. Classification Performance
  15. Interpretability Analysis
  16. ...and 9 more sections

Figures (8)

  • Figure 1: Overview of the IQNN-CS pipeline.
  • Figure 2: Accuracy and loss curves for IQNN-CS. (a) accuracy and (b) loss for Dataset 1; (c) accuracy and (d) loss for Dataset 2.
  • Figure 3: Saliency maps for a selected test sample. Dataset 1 shows concentrated attributions; Dataset 2 appears more diffuse.
  • Figure 4: t-SNE projection of quantum embeddings. (a) Dataset 1 yields a clear separation; (b) Dataset 2 shows an overlap.
  • Figure 5: Prediction confidence vs. occluded features. Left: Dataset 1 shows sharp drops; Right: Dataset 2 decays smoothly.
  • ...and 3 more figures