Machine Learning for Microsoft Azure is a cloud analytics service. It allows you to create analysis models to be deployed or changed according to need. It can also be used with algorithms that analyze data or data history in order to identify patterns to predict an event.
What is machine learning?
Before we really talk about Machine Learning in Azure (AzureML), we need to know what it is, what it is for, and its impacts. In 1959, Arthur Samuel defined machine learning as “a field of study that gives computers the ability to learn without being explicitly programmed.” Tom M. Mitchell described it this way: “It is said that a computer program learns from experience with respect to some kind of task T and performance P if its performance P in T tasks, as measured by P, improves with experience E,” demonstrating that machine learning is done in an operational and non-cognitive way.
How to start
Microsoft and other global companies see that machine learning can be widely and consistently used to predict trends, results, and future behaviors. Currently, we can see the use of machine learning in web searches and in the offering of products that appear in ads after people complete searches for those products.
Microsoft has https://studio.azureml.net, which is a collaborative tool that can be used to create, test and deploy solutions with ready-made templates or those created by the user. It is in this portal that data science, predictive analytics, and its data will work together to generate your final product. When accessed, the portal shows a screen with modules that can be chosen according to the needs of your project. For the demonstration in this article, I’ve chosen Free Workspace.
If you are not registered, basic registration will be required before proceeding. After registration, you can then enter the main screen of the studio (which is very similar to the Azure environment). From this screen, you can initiate your projects and upload files of the data that will be analyzed.
Understanding menu options
- PROJECTS: collections of tests, datasets, notebooks, and other resources that represent a single project.
- TESTS: Tests you have created and executed or saved as drafts.
- WEB SERVICES: Web services you have deployed from the tests.
- NOTEBOOKS: Jupyter Notebooks that you have created.
- DATA SETS: Data sets that you loaded in the studio.
- TRAINED MODELS: Models that you have trained in tests and saved in the studio.
- SETTINGS: A collection of settings you can use to configure your account and features.
At the bottom of the left-hand screen, click +New to view other possibilities such as tests, project templates, and examples that you can use.
- Dataset: Upload a new dataset from your computer.
- Module: Upload a custom module or use an example provided by Microsoft.
- Project: Create a new, empty project.
- Experiment: In this module, you can create a blank experiment or click on View More in Gallery and filter through the various options. You can also search for the most diverse experiments so that you can study and test the features.
- Notebook: You can upload a notebook from your local computer or use an example from Microsoft by clicking on View More in Gallery.
When searching for an experiment in the gallery, you will be asked to copy it and choose the region in which workspace will be made available. You can then proceed!
For my example, the restaurant ratings experiment was chosen. It uses Azure’s Machine Learning to identify, based on customer options, what dishes they are most likely to order. This information can help the waitstaff make stronger recommendations.
After choosing your experiment, you can start testing and learning more about how machine learning works through the available examples. A good start is taking a deep look at each item in the experiment, such as Saved Datasets and Data Transformation.
Machine Learning in Azure is very user-friendly, even for beginners. The platform has support and examples of how it can be used, as well as documentation, training and videos. Note that new professionals in this market have to understand programming, data manipulation, and math in order to act as a data scientist.