This Fintech company developed and optimized a gradient boost algorithm using SAS Enterprise Miner to identify the creditworthiness of their customers. Unfortunately, they could not implement the SAS program into their production environment. As a result, the company needed to convert the SAS code into Python to integrate it into their enterprise production system.
SCALESOLOGY IN ACTION
This Fintech company provided Scalesology an anonymized test data set that included the input data to their model as well as their resulting predictions. They also included the optimized XGBoost model score code generated by SAS Enterprise Miner. This complex gradient boosting decision tree contained over 11,000 lines of SAS code.
Scalesology converted this program using Python Pandas data frame data structures and if/then logic. Specifically, we rewrote:
The data ingestion and transformation process which parsed the input data from the preliminary credit attributes into bins.
The scoring algorithm that consisted of hundreds of decision tree leaf nodes .
Within a short time period, the Fintech company had Python code that they could deploy in their production environment. As a result, their gradient boost algorithm can be deployed within their enterprise application providing real-time credit risk evaluation of their customers.