Exploring Azure ML Studio(Classic) - Basic steps to get started with Azure ML Studio(classic)
To develop a machine learning prediction model, we collect/gather data from one or more sources, explore, transform, and analyze that data using various data manipulation and statistical functions, and generate the results. We then modify various features/parameters and their functions, until we are satisfied that we have built and trained an effective model.
Microsoft has given us an interactive tool called Azure ML Studio classic. As per Microsoft, Machine Learning Studio (classic) is a drag-and-drop tool that one can use to build, test, and deploy machine learning models. Studio (classic) publishes models as web services, which can easily be consumed by custom apps or BI tools such as Excel.
Azure Machine Learning Studio is a Web-based integrated development environment(IDE) for building and operationalizing Machine Learning models/workflow on Azure.
We just need:
Azure Machine Learning Studio account (you can create a free account for a limited time on Azure).
To start using Azure ML Studio classic, click on this link https://studio.azureml.net/
On the home page, sign up with your Microsoft account/school account/work account. Once you logged in, you will see the workspace like this:
The following are the main features of ML Studio classic:
Can manage datasets
Can create an ML model with drag and drop with low code options
Can add R and Python code module
Can manage a trained model
Publish model as web services
There are many components in the ML Studio workspace:
Create and manage projects. We can have all the related experiments under the project.
The experiment is the main component that provides the canvas to interactively work with the datasets and all other modules like various models. An experiment consists of datasets that provide data to analytical modules, which we connect together to construct a predictive analysis model.
When we select experiments, we see the following screen.
Basically, all the saved experiments will be listed and a preview of the experiment will be shown in the rightmost pane. To create a new experiment, click on the “New” button that is there in the bottom left corner highlighted with a red box.
On the “New” button click, the following screen will pop up, wherein we have to select the “Blank Experiment” icon to start with the new experiment.
The blank experiment canvas will look as shown below:
The left side panel has all the modules and components that can be used to build the model and play around with various options.
You may want to review Microsoft documentation on creating the first experiment on ML Studio classic to understand more about experiments in ML Studio classic.
Once the ML model is ready, we can deploy it as a web service right from Machine Learning Studio (classic). Using this component, the training experiment can be converted to a predictive experiment, and then publish as a web service so that the built model can be accessed by others or any web applications can use it.
A dataset is a set of data that has been uploaded to ML Studio (classic) so that it can be used in the modeling process. There are many sample datasets included with ML Studio (classic) for us to experiment with, and we can also upload more datasets as we need them from various resources.
The list of available datasets in ML Studio can be found here.
A sample experiment might look like this:
Once the experiment is completed, we can click on the “Run” button (highlighted green box) to run the experiment. Each and every step is executed and once it is finished, the results can be viewed by right-clicking on the small circle at the bottom of every module.
Setup as Webservice
The experiment can be set up as a Webservice by just clicking on the “Setup Webservice” button (highlighted blue box).
To know more about this, you may want to review this tutorial.
Publish to Gallery
Azure AI Gallery is a community-driven site for discovering and sharing solutions built with Azure AI. The Gallery has a variety of resources that we can use to develop our own analytics solutions.
Once the model is evaluated, we can publish it to Gallery so that it will be available for others to use.
To do that: click on the “Publish to Gallery” button at the bottom of the screen highlighted with the red box.
This link provides the details of how to publish to the gallery in detail.
This was a very brief overview of Azure ML studio classic, which will be useful for a beginner like me to just get started. But this alone will not be enough. The Microsoft documentation links in this blog will be really helpful to create and deploy ML models.
Some of the useful links to Microsoft documentation for Azure ML Studio classic: