BUSINESS INSIGHTS

Mar 30, 2015

Azure Machine Learning: An Overview

Mike Cornell Posted by Mike Cornell

On February 18, 2015, during the Strata Conference in San Jose, Microsoft announced the general availability of its cloud-based, predictive analytics solution, Azure Machine Learning, which had been in Preview since last July. Microsoft also announced the release of several new pieces of functionality and a new free tier of the cloud service. 

Today’s Challenges with Advanced Analytics Projects

Predictive analytics, machine learning, and data science are all hot topics at universities, tech conferences, and corporations around the world. Many companies realize that in addition to traditional BI, there are huge opportunities for competitive advantage by using historical and even real-time data to accurately predict future events or gain insights not easily detected by looking at a bar chart or scatter plot.  What_is_Azure_ML.png

Understanding the opportunity that advanced/predictive analytics provides is often obvious, but implementing these sorts of solutions often raises the same set of challenges:

  1. The need for people with a very specific, very expensive, and very hard-to-come-by skillset:  statistician + application developer.

  2. The tools and methods to develop machine learning applications are often very expensive and have historically led to projects that take months to develop and implement.

  3. Models are typically developed and consumed in a very specific programming language for a very specific application and are not regularly available for use across the organization. 

Azure’s Machine Learning service answers many of these problems with their latest release.

 

What is Azure Machine Learning?

Azure ML is an end-to-end, cloud-based, advanced/predictive analytics platform.  It caters to organizations, users, and data scientists of all skill-levels and experience. Azure ML allows users to import training data, build, train, and deploy machine learning models, and even predict outcomes and cluster data all from a simple web browser.

In addition, users can easily import training data from a number of sources including flat files, Azure BLOB storage, Hive queries, Azure SQL Databases, website lists, and more.  They can then use the simple, familiar drag-and-drop interface to create common workflow tasks for prepping data, selecting features, and training, scoring, and comparing models.

Azure ML offers a vast collection of built-in data transformation tasks, robust and proven Microsoft Machine Learning algorithms, and support for frequently used data science programming languages like R.  Once a model has been developed, it can easily be deployed and published in the form of a web service hosted on Azure.  Once deployed, the model can then be accessed from almost anywhere including custom applications, web sites, Azure Data Factory, Excel and Power BI, and more for use in predicting and clustering new data as needed.  Users and companies alike can also monetize the models they create by publishing them to the Azure Machine Learning Marketplace

Get an inside look at it in this video:

 
 
 
 
 
5:22
 
 
 
5:22
 
 
 
 
 
 
 
 
 
 
 
Wistia video thumbnail - Azure ML-Overview with Power BI by BlueGranite

Thanks for reporting a problem. We'll attach technical data about this session to help us figure out the issue. Which of these best describes the problem?

Any other details or context?

Cancel
message
 
 
 
 
 
 
 

New Functionality

With the announcement of Azure ML’s general availability, Microsoft also announced several new features with the release:

More intuitive web service creation With a single click, users can take a “training model” and turn it into a “scoring model.” Azure ML will also automatically suggest/create the input and output points of the web service model. Finally, once created, Azure ML presents an Excel file that can be downloaded and used to interact with the web service for inputting features and outputting scores/predictions.

The ability to train/retrain models through APIs – Developers, users, and customers can now periodically retrain a deployed model with their own, fresh data programmatically through an API.  This means as patterns in data change over time, or if new customers have their own ‘different’ data, they can easily connect to the existing model, retrain it using the new data, and begin predicting with the updates to the model.

Python support – The commonly-used programming language Python is now fully supported in Azure ML.  Adding custom Python code is as simple as dragging the “Execute Python Script” workflow task into the model and inputting the code directly into the dialogue box as shown below.  Azure ML even provides hints as to the inputs and outputs of the script. With the support of Python, this means a user can now integrate Python, R, and Microsoft ML algorithms all into a single workflow.  That’s a pretty heavy arsenal!pythonsupport.png

Learn with terabyte-sized data – Users can now connect to and build predictive models using “Big Data” sets with the support of a new feature transformation capability called “Learning with Counts.”  More information about this new feature that uses mapreduce functions in Azure HDInsight can be found here.

Azure ML Community Gallery – Microsoft has introduced a new community-driven site that serves as a great place to find tons of sample experiments and to learn from others who have already been working with the service.  Users can even share their content on common social media outlets such as LinkedIn and Twitter.

New Pricing Structure + Free Tier

Finally, the announcement from Microsoft also included the introduction of a new pricing structure for Azure ML including a free tier.  The comparison of the 2 tiers is presented in the table below from Microsoft’s site.

In general, the free tier is great for smaller projects, as the user can work in ML Studio with up to 10 GB of data and can deploy ‘staging’ web services to use/test with.  The standard tier brings with it all of the support and scalability of Azure.  Users can work with virtually unlimited data sets, access even more data sources, deploy web services to production with more scalable/configurable performance, and ultimately harness the full capabilities of the Microsoft cloud.

Where do We Go From Here…

In Microsoft’s announcement of the general availability of Azure ML, Joseph Sirosh, Corporate Vice President of Information Management & Machine Learning at Microsoft, says that Microsoft wants to

“… eliminate the heavy lifting involved in building and deploying machine learning technology and make it accessible to everybody.”

At Blue-Granite, we are very excited about the latest release and availability of Azure ML.  We believe that Azure ML solves many of challenges that today’s advanced analytics projects present, and see this game-changing service as a statement of Microsoft’s commitment to being a major contributor in the advanced analytics space.

Want to find out more about how Azure ML can fit into your advanced analytics strategy? Contact us today to learn more or click here to view three hands-on labs for Azure ML to get you started. 

 

New Call-to-action
Mike Cornell

About The Author

Mike Cornell

Mike Cornell is a former BlueGranite employee. He is passionate about helping to solve business problems of varying size and complexity using data and analytics. Mike's specializations include big data platforms, cloud data platforms, advanced analytics, and data visualization and exploration. His technology interests include the Azure Data Platform, Hadoop Data Platform, Spark, R and Python for data analysis, Power BI, and SQL Server. Check out Mike's blog at http://www.datamic.net .

Latest Posts

New Call-to-action