Hitachi Capital Consumer Finance wanted to understand the uplifts that may be possible through using a deep learning approach. The Consumer Finance division appointed Jaywing to explore the potential to significantly improve its existing application credit scores using Jaywing’s AI modelling technology, Archetype.
“Archetype solves the black box problem in credit risk and has given us food for thought in how we approach our modelling activity.”
Jaywing processed data from Hitachi’s recent scorecard development, using Archetype to replace the traditional modelling steps. Using the software, Jaywing was able to apply appropriate constraints which reflected how Hitachi’s modellers would apply common sense rules to the inputs. These rules meant that the system would only generate model outputs adhering to Hitachi’s expectations, making models explainable to regulators and customers alike – a key differentiator compared to other Neural Network-based approaches.
- Archetype demonstrated an impressive uplift of 7.2% compared to an optimum linear regression model, showing that the Archetype model had the potential to predict and prevent more bad debt or to increase the number of customers taken on without increasing bad debt levels.
- On a smaller sub-prime portfolio, an uplift of 11% was seen, from a lower baseline.
“Not only did Jaywing promise uplifts through the Archetype software, they delivered them too."
Nick Gibbs, Head of Commercial & Strategy at Hitachi Capital Consumer Finance
Jaywing went on to produce a production-strength model for Hitachi Capital to deploy independently.