Optimizing Models

Operations teams should simplify and optimize the workflows that are created in the development environment by the analytics team. Over time, the operations team can use the collaborative environment to encourage and enforce best practices, by creating template workflows and custom operators that encapsulate standards and optimizations.

Here are some best practices when optimizing your workflow for production:

  • Remove non-essential operators such as charts, summary statistics, accuracy tests, and unapproved models.
  • Consolidate multiple operators into one (for example, successive variable operators).
  • Perform pre-aggregations and filters as early as possible.
  • Perform pre-aggregations and filters at the source of the data.
  • Use subflows to standardize common transformations.
  • Use custom operators to streamline complex operations into a single step, and to standardize common operations. (See TIBCO® Data Science Team Studio Development Kit for more information.)

You can score, deploy, and manage your models in one place. You can also automate job runs and integrate them into your development process from the command line using our APIs.

  • Batch Model Scoring
    To batch score new data in a workflow, first export the desired model in Team Studio Model Format using the Export operator.
  • Real-Time Model Scoring
    The Export operator can also export in PMML and PFA formats.
  • Code Generation
    ETL workflows on databases generate SQL logs that can be ported to other database environments.
  • Model Management
    Starting in version 6.3, Team Studio provides advanced features for model governance and deployment to real-time scoring services. In the Administration section, you can modify the deployment targets to which Team Studio has access.
  • Workflow Scheduling
    Your workflows and models can also be scheduled in a job. Jobs can run a variety of tasks, ranging from simply executing a workflow to refreshing, scoring, and deploying models.
Related concepts