AI Powered Credit Scoring Algorithm - Redefining Credit Risk
OUR USE CASE JOURNEY
A consumer financing company wants to streamline its financing eligibility time and automate its decision making:
Time. When a loan is requested, loan officers have to perform inquiries in various locations which require effort and take too long to process the vast number of applications streaming in.
Cost. Due to the large number of loan officers hired to manage the large amounts of inbound applications, the cost becomes too high to sustain quality and speed of processing.
Human Bias. Each loan officer can have a different opinion on credit limit & terms for the same application, which leads to high bias and unmanaged risk.
The company collects the following data to allow the interrogator to double check and set credit terms and limits.
The aim is to avoid biases in future analysis & streamline the decision-making process to take a fraction of the original time while minimizing cost.
Data Features. To accurately predict the eligibility and credit terms for each applicant, all the previous features regarding applicants were taken into consideration to determine the hierarchy of significant.
Model Design. The machine learning algorithm was designed to give applicants a score of 0 to 100 while measuring correlation between score and customer demographic data.
The Algorithm. By blending all the data sources alongside demographic and macroeconomic data, a machine learning solutions was used to provide instant credit scoring with minimum bias and maximum accuracy.
Instant Scoring. The credit scoring algorithm provides instantaneous decisions with 100% confidence for 40% of new applicants.
Minimizing Uncertainty. The other 60% are asked to provide only a few documents or are subjected to minimal investigation of a handful of factors only instead of full in-depth investigation.
Levels of Intelligence. All the intelligence layers serve to give instant decisions, set credit limits and terms and continuously change with new inputs for recurring applicants.