I was a little bit confused on my metamorphosis from the role of a data miner to data scientist. For years I was using data mining tools and techniques both proprietaries like SAS Enterprise Miner and Clementine (now SPSS Modeler) and open source counterparts like R and scikit-learn. My customers and I were happy with SAS' EM excellent capability on pattern finding and predictions, R's availability of high quality algorithms and visualization packages and easiness Numpy and SciPY with sci-kit learn. But both of us were not that happy with the deployment and maintenance procedures required of these software packages.
It is true that most of the data miners were not that much concerned about the deployment and maintenance of their own projects. They were more interested to finding new patterns, experimenting new data mining algorithms, creating different optimization strategies and complaining about the data quality to the ETL team! Choosing an end to end product solutions under one umbrella starting from data ingestion and going via data management tools, data stores, and ML/DM software packages to dashboards and other business workflows required heavy investments. I think, Microsoft's new Cortana Analytics Suite is trying to fill this gap.
Cortana Analytics is centered on Azure Machine Learning, supported both by wizardy R and energetic Python. CEP and stream analytics supported by Azure Stream Analytics and widely used big data open source tools - Spark and Storm bundled on top of HDInsight, Microsoft’s Hadoop distribution. Azure Data Lake and an elastic data warehouse based on SQL server is available as an alternative to Hadoop. Azure Data Factory can create an ETL style orchestration, data cataloging via Azure Data Catalog and Event ingestion via Azure Event Hubs. Dashboards and visualization can be via Power BI or R (ggplot2, ggviz) or even with IPython or Zeppelin notebooks. Azure ML comes with a lot of pre-configured solution templates for business scenarios like recommendations, forecasting, churn, etc.
I spend a lot of time for creating and validating multiple machine learning algorithms to choose the best model and deploying the selected solution. Microsoft make this task very easy in Azure ML. Even ML models can be published outside as web service in R, C# and Python, so it can be consumed by different LOB applications making them more smart and intelligent.
Everything on cloud, Azure. I wish a standalone version of Azure ML - but looks like they have no plans for this. For on premise, integration of R (recently MS acquired Revolution Analytics) with SQL Server 2016 give new promises for SSAS data mining.
I am not leaving SAS, they are gold standard for analytics - trusted and time proven; but will watch an eye on Cortana Analytics Suite.