Business solution examples

CASE 2

AI Delinquency Prognostic Analysis Robot

Can AI identify people who may be delayed?

Case overview

There was a request from a financial institution that was trying to find customers who may be delayed (delinquency prognostic analysis). We solved the problem through the analysis of big data using AI.

Background / Challenges

  • We want to improve the work efficiency in accounts receivable.
  • We want to analyze indications of customer delinquency by AI.
  • We would like to reduce the cost of debt collection.
  • We would like to develop new low-interest, stable financial products and services.

Results / Outlook

We have achieved a very high prediction accuracy and AUC score. We were able to clarify important indicators to predict the possibility of delinquency.

Our Solution

Learning with Deep Neural Network
Learning with Deep Neural Network

In addition to information such as customer's age, occupation, payment information, credit card cashing limit and the like, POS information can be used to see what products the customer purchased and at what frequency, which is then input to the deep neural network and analyzed to predict those who may be delinquent.

Prediction accuracy of delinquency indicators
Prediction accuracy of delinquency indicators

With AI prediction, a prediction accuracy of 98.5% is achieved. Also, the AUC score (the closer to 1, the higher the analysis ability) is at a high level with a score of 0.946.