Data scientists are producing more analytical models than ever – but there's been no corresponding increase in business value. That's because few models make it out of the lab and into production. SAS Open Model Manager can change all that by helping you operationalize your open source models and put your data to work for smarter, faster decisions.
Easy model governance.
Centrally catalog and manage different types of models for greater security and reliability. SAS Open Model Manager enables you to easily understand the definition, properties and function of your analytical models − including who is testing, validating and approving different models – and fosters collaboration between IT/DevOps, data scientists and the model validation team.
Key Governance Features
- Ensure proper model registration, access and administration – including backup and restore capabilities, overwrite protection and event logging – with secure, centralized storage for different types of models.
- Search, query, sort and filter models by the attributes used to store them – type of asset, algorithm, input or target variables, model ID, etc. – as well as user-defined properties and editable keywords.
- Add general properties – model name, role, type of algorithm, date modified, modified by, repository location, description, version and keywords (tags) – as columns to the listing for models and projects.
- Access models and model-score artifacts using open REST APIs.
Simplified model deployment.
Deploy your analytical models with operational systems or processes at scale in a way that's repeatable and traceable. Simplified, automated publishing and scoring steps give you the flexibility to operationalize models with just a few clicks – both in batch and real time – in different operational environments.
Key Deployment Features
- Combine open source models, including Python, in the same project for users to compare and assess using different model fit statistics.
- Set up, maintain and manage separate versions for champion and challenger models.
- Define test and production score jobs for analytical models, including Python, using required inputs and outputs.
- Publish models to batch/operational systems (e.g., in database, in Hadoop, Spark).
Model performance monitoring.
Avoid model decay and revalidate the business value and impact of models in production by monitoring model performance to determine the impact of changes in market conditions or customer behavior, new data or concept drift.
Key Monitoring Features
- Monitor the performance of models with any type of score code. Produce performance reports – including variable distribution plots, lift charts, stability charts, ROC, K-S and Gini reports – for champion and challenger models using performance-reporting output result sets.
- Evaluate the need to retrain, retire or create new models using built-in reports that display the measures for input and output data and fit statistics for classification and regression models.
Containerization for application portability.
Containerized analytics model management captures all of the environmental dependencies for your analytic workload, providing a portable, lightweight image that you can deploy anywhere. Containerization also enables you to deploy models in efficient, cost-effective cloud-based systems. And you can deploy models at scale by clustering and scheduling container processing in a distributed environment.
Key Containerization Features
- Easily deploy SAS Open Model Manager components within container-enabled infrastructures, including Docker and Kubernetes, which are often run in the cloud.
- Publish analytical models to runtime containers with embedded binaries and score code files. Promote runtime containers to local Docker, AWS Docker and Amazon EKS (Elastic Kubernetes Service) environments.