SQLServerCentral Article

Data Mining Part 39: Azure Machine Learning vs Microsoft Data Mining



We talked about the Microsoft Data Mining enemies in the chapter 32This time, we will talk about Azure Machine Learning. Is it the same than Microsoft Data Mining? What is the difference?


This time you do not need to have SQL Server Installed nor the Azure Portal. Just relax, go to a comfortable sofa and take a good sandwich and a soda to read this article.

Getting Started

Microsoft Azure is a set of services in the Cloud like Web services, Web Pages, Virtual Machines, Machine Learning Services and other services. This time, we will talk about Machine Learning because it offers similar capabilities than the SSAS Microsoft Data Mining.

What is Machine Learning (ML)?

According to Arthur Samuel (a famous pioneer on the machine learning field), it is a field of study that gives computers the ability to learn without being explicitly programmed. Machine learning is when the machines learn things by themselves like playing chess (and improve each time they play), OCR (Optical Character Recognition) and many other fields. To learn, it uses algorithms like the Cluster, Decision trees, Neural Networks and other algorithms that we learned with Microsoft Data Mining.

Data Mining is part of the machine learning world. In Data Mining, the machine learns to find patters. Machine learning includes the Data Mining as part of the capabilities.

What is Azure Machine Learning?

It is an analytical tool in the cloud that can be accessed by a smartphone, a tablet or any other device that connects to Internet. This tool is used to analyze information and find patters. 

Is Machine Learning replacing Microsoft Data Mining?

Yes and No. Machine Learning can be used to replace Data Mining or vice versa. Data Mining is simpler. Very easy to learn and very practical. Machine Learning is a more complex and complete tool. Machine Learning contains more algorithms, it can be integrated with R, which is a very popular language these days and SQL Server will support R in the version 2016. 

In addition, people love that it is easy to create Web Services in Machine Learning and it is easy to integrate with languages like R, C# and Python (languages that are very popular in other Data Mining applications). It is recommended to use Data Mining instead of Machine Learning if you:

  • Do not want to send your data to the cloud (for security or other reasons).
  • Do not have a trustable internet connection to trust in the Cloud.
  • If you already have Analysis Services installed or you already have SQL Server installed and you only need to install the Analysis Services.
  • If the model can be accomplish by the Microsoft Data Mining capabilities.

Microsoft Data Mining is easier to use and learn. It is very simple. Machine Learning is recommended if:

  • You want to easily access to your Analysis Reports using the Web from any smart device.
  • You do not want to install SQL Server.
  • You do not want to spend money on the software (there is a free edition of machine learning).
  • You need to add complex queries and code using Python or R.
  • You need more complex algorithms that the ones offered by Microsoft Data Mining (it is possible to create and add new algorithms in Microsoft Data Mining, but it is not a straightforward task).
  • You do not want to install Data Mining tools. 
  • You want to access to your data anywhere.
  • Your Data Sources are not supported by Data Mining (ML supports more types of data sources).

What are the prices for Microsoft Data Mining and Machine Learning?

For Microsoft Data Mining you need to purchase the Enterprise or the Business Intelligence Editions. The Standard Edition is very limited. The price for the SQL Server Enterprise edition is 14,000 USD per Core approximately. The price for the SQL Server Business Intelligence edition is 8.900 USD per server approximately.

The prices for the Machine Learning services are more complex. If you want the free edition, you can only have 100 modules per experiment (an experiment is like one project and the modules are the subsection like training data, evaluating data and training the data), the maximum storage is 10 GB, The maximum experiment duration is 1 hour and you also have limitations related to the performance and API production.

The standard Machine Learning version charges 10 USD approximately per seat and per month. A seat is a workspace where you develop your experiments. You are also charged 1 USD/hour per usage of the ML Studio. In the ML Studio, you create your experiments. You will also be charged with 2 USD/Production API Compute Hour and 0,5 per 1000 production API transactions. My recommendation is to use the free version for small experiments, and once they run OK, migrate them to the standard version and execute them with the real information.


The future looks to be in the Cloud. Microsoft is dating with Linux to grow faster and in Azure there is no more discrimination to non-Microsoft technologies. It looks that the users will move from Microsoft Data Mining to the Machine Learning.

Azure has more Data Centers that Amazon Web Services and Google Cloud Services combined. 85 % of the fortune 500 companies are using Azure. The number of Azure users is increasing at a rate of 120,000 users per month approximately and there are 1,4 millions of databases in Azure. 


For more information, refer to the following links:


4.17 (6)




4.17 (6)