Finance

AI-based Cardholder Credit Scoring System

The Challenge

Our client requires an advanced model that maximizes the accuracy of default prediction, using historical activity as training data.

| The Outcome

Increased business through our predictive model detecting the risk of non-payment of existing customers and optimizing the qualification of new applicants for the means of payment.

AI-based Scoring System

How we did it

  • We provided an AI system for automatic recalculation of scoring weights and, subsequently, implemented a Machine Learning model to predict defaults.
  • We evaluated the accuracy of the ML model using a sample of test data to estimate the future default prediction hit ratio.
  • Optionally, a different date range (per user) could be selected from the default date range proposed by the system.
  • The hit ratios for customers accepted or rejected due to risk of non-payment were improved with respect to the existing weighting systems.
  • The model is frequently updated and retrained effortlessly, evaluating its performance/estimated accuracy.
  • The possibility of including new variables and customer data not initially taken into account is designed to improve the accuracy of the model.

  • Early Detection

    Reduced detection time of non-payment risk

  • +26%

    Increase of new customers

  • 1/6 Defaults

    More accurate insolvency prediction

Shall We Talk?

If you need to know more specifics or are interested in having us assist your organization, please use the form below.