Microsoft Azure Review

The Canvas flow interface brings a very nice functionality that improves self-learning.


Valuable Features

The Canvas flow interface brings a very nice functionality that improves self-learning.

Improvements to My Organization

Selecting and applying a machine learning algorithm is not a click-and-run process. In this case, clustering our data has helped us to find patterns and trends that were not visible using conventional (internal) classification.

Room for Improvement

Stronger R integration in a circular fashion (Azure->R->Azure). Nowadays, it has an unidirectional bias.

Use of Solution

I've used Machine Learning Studio for the last six months.

Deployment Issues

We've had no issues with the deployment.

Stability Issues

By using a free account for cloud services can bring you "queued" status, but normally it runs smoothly. The quota defined for free account is enough for a variety of experiments and none of the features are blocked.

Scalability Issues

We've had no issues with the scalability.

Customer Service and Technical Support

Customer Service:

As a cloud service, online chat and email service are available, but the community forums are the best place to solve issues.

Technical Support:

We have a support maintenance agreement for Windows/Office.

Previous Solutions

There are a few competitors in place, and their number is increasing, but friendliness is the strong point of AzureML, comparing to Amazon AWS or BigML.

Initial Setup

It does not involve any setup, just having an Outlook or Microsoft account.

Pricing, Setup Cost and Licensing

Using a free account leads to offline projects. For online projects those issues must be evaluated according to client side company's environment.

Other Advice

Machine learning is just one part of the whole data science cycle. Big data (streaming, video, etc.) or deep learning needs must be addressed with additional tools. But for prediction/classification this is a fantastic tool.

Below is a PCA graph generated by K-means training model and its Qlik Sense panel. The idea is to segment unlabeled data based on numerical features in order to find common patterns that can be grouped, named clusters.

Disclosure: I am a real user, and this review is based on my own experience and opinions.
Add a Comment
Guest
Sign Up with Email