Executive Summary

Effective Credit Risk Management (CRM) is the backbone of any commercial bank’s risk management framework. The credit assessment process requires the assessor to derive an opinion on the credit worthiness of a borrower using a diverse range of information such as exposures, financial performance, company developments and external business environment. The information gathering process is often manual and time-consuming.

Our PoC aims not only to help streamline the data information gathering process, but also relook at the way data sources can be organized and visualized in a more insightful manner using knowledge graph and semantic ontology.

With a credit grading recommendation engine and an early warning flag built within the system using advanced data driven methods like machine learning, we aim to generate an educated credit opinion using deep learning models (i.e. LSTM and SEQ2SEQ). Ultimately, credit risk evaluation would evolve into a process that is more time sensitive and predictive.

The positive outcome of our PoC suggests that the introduction of an engine equipped with basic cognitive ability similar to a human being would be a milestone towards in the development of explainable AI. Not only will the engine produce a predictive outcome for human consumption, it will also be able to explain how it managed to derive at the conclusion.

View original post at mizuhogroup.com/asia-pacific/singapore/about/achievements.

Read the Whitepaper (PDF/1.03MB).