SAS Viya, a "cloud-friendly" rethink of the SAS portfolio designed for self-service, has recently added Model Studio, a tool that provides a unified workflow from model development to deployment and lifecycle management. These are important pillars that SAS needs to prevail in an increasingly competitive market; the next step should be adding machine learning to the interface to provide a virtual assistant to optimize the workflow.
Turning modeling into a pipeline process
The new Model Studio web application from SAS Viya adapts capabilities from SAS Enterprise Miner, Visual Analytics, Visual Data Mining and Machine Learning, and Visual Text Analytics to provide a unified lifecycle management workflow or pipeline, all within the same environment, removing the need to boot up separate tools.
It starts with creating or harvesting models. Users can create models in Model Studio or create and import them from Jupyter notebooks. The models could also be written in open source languages such as R or Python, and then scored in SAS. Model Studio can automatically tune parameters fed into machine-learning models ("hyper parameter tuning"). Users can also drop into a data preparation session to clean up the data sets, then move to deployment by dragging and dropping different models onto different data sources. Capabilities borrowed from Visual Analytics enable users to then visually edit the results, dragging and dropping columns, splitting columns, and trimming white space for presentation. Users can then deploy the models as visualizations, or as streaming data via SAS Event Stream Processing, as a REST endpoint, or run inside of data sources, such as Hadoop, with the use of SAS in-database technologies. A set of best practices is available for managing the modeling pipeline from cradle to grave. What's missing is embedded machine learning that can recommend best practices at each step.
Besides Model Studio, SAS has extended its established Model Manager product to the Viya portfolio. This builds a bridge for existing SAS customers to take advantage of Viya's analytic capabilities by incorporating analytic models developed through the Viya environment into their existing workflows. The extensions to SAS Viya come as start-ups are emerging through the woodwork with a variety of collaboration and lifecycle management tools that harness the fruits of the open source world.
"Viya provides new front door to SAS analytics," IT0014-003116 (April 2016)
"SAS Data Loader for Hadoop points to future convergence strategy with Viya," IT0014-003188 (December 2016)
Tony Baer, Principal Analyst, Information Management