"A neural network more closely mimics the way humans think and reason, whereas linear models are more dogmatic — you’re imposing structure on data as opposed to letting the data talk to you."
~ Eric VonDohlen , VP of Enterprise Business Intelligence and Analytics at ICW Group
Traditionally, the biggest advantage of AI-based credit scoring systems is that they can unearth hidden relationships between variables that are not always apparent to traditional credit scoring systems which look at one variable at a time.
According to Dr. Stephen Coggeshall, who is an analytics and data science veteran with 11 years of experience in the field, it is hard for traditional
credit scoring systems to unearth those hidden relationships because a lot of data pre-processing and expert knowledge is required before one can even attempt to find those non-linear relationships. Something that an AI-based credit scoring system can do with
ease, and with minimum effort.
AI-based credit scoring models provide a more nuanced evaluation of data and can consider data that would not seem relevant or even included in a traditional credit scoring model. AI can provide rules that are very complex and in-depth, as opposed to traditional
credit scoring models that use very simple rules, and often end up rejecting borrowers who are credit-worthy. Also, a self-learning credit scoring model can continuously improve itself as new data gets fed into the system. A feature that traditional credit
scoring models do not possess.
The Predictive Process
- Traditional credit scoring systems make assumptions and test based on historical data to predict future credit-worthiness
- A self-learning AI analyzes data, learns from it, improves itself and provides predictions at a scale and depth of detail impossible for a standard credit scoring model.
Ease of Re-calibration
Sooner or later, traditional models become outdated. Every time that happens, we have to call in experts, invest substantial amounts of money and time to reevaluate the model. AI-based scoring is dynamic in the sense that it can update itself. Also, when
alternative data becomes available, it can be retrained to create challenger models.
During the first few months (the duration can be less or more depending on how much data the FI can accumulate each month), use of AI in credit scoring will provide only an incremental boost in spotting customers over traditional credit scoring. However
as more data is absorbed by the system, an AI that uses advanced techniques like ensemble learning and continuous improvement, will become far more accurate than and efficient at spotting credit-worthy customers that traditional credit scoring models are bound