Intro
You don’t have to be working in the IT/CS these days in order for your ears (or eyes) to pick up the term Machine Learning (ML) or Data Science, since these are among the biggest tech trends nowadays. Machine Learning, being the subset of Artificial Intelligence (some of you might disagree with this) is a concept ubiquitous in our daily lives.
If you are using a Personal Assistant such as Cortana, Siri or Google Now– as soon as you wake up, you can get information such as which traffic route to take to work, based on previous trends and patterns in traffic, or if you’re using any of the Fitness Trackers – it can learn your sleeping habits and pick an ideal time for waking up. Fast-forward through the day – as you are surfing the web – big tech giants are collecting your information and by comparing your habits to other people and their interests – and it can target you surprisingly well with ads and recommendations that suit your needs. After long working day, in not so distant future, most of all us will have a large number of sensors laid out at home, sending data to Cloud in order for a ML system to pick up our evening routines and automate heating systems, lighting, and we’ll see what else. Or – if you are inpatient – take a look at I.T. (2016) movie :).
This is not something revolutionary new, we have these concepts around for quite some time now. And those are so omnipresent that the industry needs more and more people working with ML everyday to fulfill the business needs. But Machine Learning is hard. The field is still young, research papers are published daily and it’s a combination of Statistics and Probability, Mathematics and Computer Science for the most part. In order to start from scratch and understand the underpinning of various algorithms used, and optimize them and scale effectively, you’ll need a scientific background and even probably a PhD in some of the quantitative fields.
Luckily, the ecosystem matures and we see a trend of democratization of Machine Learning and this is where Microsoft and their research team is brining a new way for developers and engineers without a ML background to take advantage of breakthroughs in the field. We’ll see how we can jump the train and be a part of the revolution brought by Microsoft and their Azure cloud platform.
Azure Machine Learning (ML for All)
Azure Machine Learning Studio is a GUI-based integrated development environment for constructing and operationalizing Machine Learning workflow on Azure.
Let’s go through the basic workflow of Azure ML:
![]()
https://docs.microsoft.com/en-us/azure/machine-learning/machine-learning-what-is-machine-learning
The basic cycle is as follows: Collect -> Analyze/Model -> Consume.
The first, and probably most time consuming step is to collect the data you need in order to train your model. We’ll talk more about this later, but the point is data is often unorganized, incomplete or even conflicting. Fortunately, Azure ML Studio has a large number of tools to help you with that. Once we get our data right in sense of magnitude and correctness, we need to select the proper algorithm for model training. There are both supervised and unsupervised algorithms available from the regression, classification, clustering and anomaly detection families. Finally, after you pick your algorithm you need to evaluate and test your model. Usually, you’ll split your collected data into training, validation and test sets which you will use to train and test your model, respectively. Again, Azure ML Studio gives you support for that, and we’ll see this in one of the coming blogs.
The final result is – you have your (hopefully) well trained model, which you can use to predict outputs based on inputs that were not in your original data set. And this is where Azure ML Studio really shines – with a couple of clicks you can add intelligence to your (mobile) apps, websites or provide insights in BI tools such as Power BI or Microsoft Excel. This is done by adding a web service which acts as an interface between your model and client apps – web service can be invoked with a single request-response model, or it can be batched.
Download full diagram here.
![]()
https://docs.microsoft.com/en-us/azure/machine-learning/machine-learning-studio-overview-diagram
Start Using Azure Machine Learning Studio
The only thing you need in order to use Azure Machine Learning Studio is a Microsoft account. You get:
- Free access that never expires
- 10 GB storage
- R and Python scripts support
- Predictive web services
What will follow
So, go ahead and create an account on https://studio.azureml.net, and the next blog will follow with end to end instructions how can we create and integrate data intelligence in our client apps. Stay tuned!