This research was funded through a University of Edinburgh Major Initiative Fund grant held by the Edinburgh Centre for Financial Innovations. The data and computing facilities were provided by NatWest, Smart Data Foundry, and Edinburgh Parallel Computing Centre.

Building blocks are overlayed with digital squares that highlight people living their day-to-day lives through windows. Some of the squares are accompanied by cursors.

Abstract

This study presents an explainable self-attention framework for predicting individual financial vulnerability using large-scale longitudinal banking data. While achieving strong discriminative performance (AUC-ROC ≈ 0.90), the model prioritises transparency through outcome-aware feature attribution analysis across all prediction classes. Our results reveal that high-risk classifications are driven by asymmetric “risk activation” triggered by financial and employment volatility, specifically income variability and payment irregularity, rather than static demographic factors. Analysis of misclassifications indicates that errors stem from structural limitations in evidence aggregation rather than random noise. By linking predictive outcomes to underlying feature-level logic, this work provides a scalable approach to high-stakes automated decision-making in financial risk assessment.

Read the full report (PDF format)

Explore related research outputs: