Tabular Image: Turning financial tables into images for smarter credit scoring
Junhao Liang, Xingjie Wei and Barbara Summers have had a new paper published.
A new paper by Leeds Univeristy Business Schools Junhao Liang, Xingjie Wei and Barbara Summers presents Tabular Image, a data-transformation technique that converts tabular financial data into compact images so it can be analysed by modern deep neural networks.
This paper created a novel data transformation technique that embeds crucial credit scoring measures, weight of evidence and information value, directly into compact, resilient images. The method demonstrated state-of-the-art predictive performance, with particularly notable improvements on larger datasets.
This overall opens new possibilities for leveraging advanced machine learning techniques in financial risk assessment and provides a flexible framework that could be adapted for tabular datasets across various domains beyond finance.
