AWS 기반 SAS Viya로 모델 모니터링 자동화…보고 시간 단축과 모델 거버넌스 강화 아프리카 최대 금융기관 중 하나인 Absa 은행은 500개 이상의 신용 리스크 모델을 운영하고 있었습니다. 이 과정에서 모델 모니터링과 보고에 최대 4주가 소요되는 문제를 해결하고자 했습니다. 높아지는 고객 기대와 규제 요구에 대응하기 위해 은행은 데이터 및 AI 부문
Most machine learning models produce a probability, but many times logic is applied to that prediction to produce a decision. That last logic step often lives in a downstream script disconnected from the model it depends on, easy to lose when the model is refreshed. Using the home equity (HMEQ) dataset, this post walks through a practical alternative in SAS Model Studio. A SAS Code node placed after the modeling node weights the predicted default probability by the requested loan amount to produce expected loss in dollars, and the new Model Registration node (2026.05) accumulates that logic into a single model registered in SAS Model Manager. The result is a model and its decision logic captured as one governed, versioned artifact, so whoever scores the model gets the decision-ready output computed the same way every time.
Open-source technologies have become a standard part of modern analytics, data science and AI. Organizations across regulated industries are adopting tools like Python to accelerate innovation while building more flexible analytics environments. But adopting open-source technologies introduces a new challenge. As organizations modernize, they must also maintain the governance, transparency
