Case study
A financial services firm wants to update its credit scoring system because the existing model is based on traditional statistical models and is quickly becoming outdated. Given the influx of data and complexity in the market, their old system has become less effective. The firm is looking for a more sophisticated model that would handle the large volumes of data and the various data types and sources, and could also improve on predicting credit risk. If the new model could reduce the rate of defaults (primary goal) and improve creditworthy customer approval rates (secondary goal), the investment would be successful. Here’s why an MLP was chosen as the favorite among the DL models tested by the AI PM team:
- MLPs can integrate easily with existing systems and handle complex, diverse datasets. So, it was a practical choice because it was integrated smoothly with the firm’s existing infrastructure.
- MLPs excel at modeling non-linear interactions...