Dmitry Teplyakov: Machine learning for non-programmers


Present days we can often meet such phrases as “Big data” and “machine learning” . Gartner company pointed comprehensive analytics as one of five key IT trends of 2015. Data analytics topics are discussed at every IT conference. If you are not using these technologies now, then you are behind the modern tendencies.

But there is another opinion. Some specialists reckon that all these big data and machine learning directions got too much of attention. That following the fashion different companies implement these technologies without clear understanding of how to use them. And asking any system integrator usually leads to getting a proposal with extremely high figures and terms.

What to do if you just want to test these systems to understand do you really need machine learning?  But you don’t have software developers and NASA-like budgets? To do it, you can use existing and free tools which can give you a chance to create machine learning models with your real data and to get predictions without development skills. Microsoft Azure ML Studio is one of these tools. It allows you to upload csv or tsv data, draw a model with visual editor, choose needed algorithm and finally to publish a model and to connect Exel to this model. There is more detailed algorithm:


Sample of the final model at visual editor:


AS you can see, there is nothing difficult at this system. You only need basic knowledge in statistics and machine learning to choose correct model type. Even if you don’t have such skills, you can get them through the Internet. There are a lot of free courses at Courcera, EdX and other online education systems.

After opening your saved project at Exel, you can input needed data and receive predictions at the same page. There is the example:



At the left side you can see a set of options with your source data. At the right with green header – copies of source data and two columns of predictions. Predictions depend of your algorithm. At this sample there are mean and deviation of our sales prediction.

Therefore, such tools as Azure ML studio can allow any data analysis specialist to start using machine learning technologies with real data and to understand its necessity and usefulness for your business.

I’m using machine learning for our project. And in practice it’s not as easy as it seems at first glance. You must prepare your data correctly, choose correct algorithm and train your model. Otherwise you will get wrong results.

Azure ML studio is a part of Microsoft Azure cloud system. But if response time is not important, you can use this system for free. If you don’t know which model better suits you, you can find about 20 already prepared projects with data samples and descriptions which can help you better understand machine learning technologies.