12-14-2017 02:45 PM
Hello fellow data scientists! On December 12, SAS released an updated version of SAS Visual Data Mining and Machine Learning on SAS Viya. Now, this was no ‘ordinary’ release. For the past few years, we have been very busy here at SAS re-imagining the data science experience, but with one wrinkle. Why should data scientists have to use a different environment than business analysts? What if you were a pure python or R programmer and was a member of a team of SAS users? What if you wanted to go end-to-end, from data prep to discovery to machine learning to deployment, all within a few clicks?
Along comes SAS Visual Data Mining and Machine Learning 8.2 – a true end-to-end analytic experience that requires no alt-tabs and adapts the experience based on the ‘business actions’ you want to undertake. You click on ‘Prepare Data’, ‘Explore and Visualize’ and ‘Build Models’. This subtle switch in experience should make your analytic experience much smoother, and more dynamic.
The integration between the ‘analytic actions’ is streamlined. As we announced in version 8.1 earlier in 2017, you can interactively build neural networks, forests, gradient boosting machines, support vector machines and factorization machines.
With 8.2, you can continue the experience by ‘creating pipelines’ from these interactive visualizations. You can drag-and-drop nodes onto your pipelines and extend your analysis. Perhaps you have a more advanced data scientist who wants to add in feature extraction to the analysis, identify anomalous behavior, or even insert code (SAS or Open Source) into a pipeline for a more robust analysis.
I haven’t touched upon my favorite part of this new release! A core theme of this product release was collaboration. Using the SAS Toolbox, you can add in a pipeline from your coworkers! You can save pipelines from existing projects and share them with users across the enterprise. SAS will also provide a handful of best practice pipelines for you to start your projects with.
Once you have generated your pipelines, you could interactively compare and assess your results. Make changes to any model. This system is not black box. With one click, you can generate a batch retrain or a simple scoring API call wrapped in Python or REST. Deploy your models into databases and Hadoop with one click, no manual rewrite.
With this new version, we also released new underlying machine learning methods. This new list includes bayesian networks, frequent item set mining, k nearest neighbors, deep learning, variable clustering along with a multitude of method enhancements.
For a full list of features, please visit: https://www.sas.com/en_us/software/visual-data-mining-machine-learning.html
Thanks for listening in. Stay tuned for even more exciting enhancements to our suite.
Principal Product Manager, SAS Visual Data Mining and Machine Learning, Enterprise Miner, and Factory Miner