Azure Machine Learning Studio

Azure Machine Learning Studio is an ML-as-a-Service platform for creating custom machine learning (ML) models. Azure ML Studio is a great tool for beginners who perhaps have some experience of consuming machine learning models and who would like to gain a deeper understanding of the training process. It offers more flexibility than the Cognitive Services APIs and an easy-to-learn development environment. The GUI does not require any programming and allows the user to concentrate on building ML models as efficiently as possible. Azure ML Studio is also a useful tool for more experienced AI developers who have a fairly simple problem at hand and need to get results quickly.

Azure ML Studio consists of two separate services: a Studio Workspace and Studio Web Services. Both of these services also include the backend computational resources needed for...

Deploying an Azure AI Gallery template

Developing models with ML Studio does not have to be done from scratch. Azure AI Gallery contains a wide selection of templates for many different scenarios. These scenarios include many common use cases for ML, such as credit risk prediction, demand estimation, and text sentiment analysis. Templates can be imported to an Azure ML Studio Workspace with a few clicks and they contain all the steps needed to produce a working ML model. Studying templates is a great way to learn about different use cases and the steps required to produce an ML model. Some templates are prepared by Microsoft, but users can also submit their own experiments to the gallery.

The template gallery can be accessed directly from ML Studio. Open the ML Studio UI (as described previously) and create a new experiment by clicking on the + New in the bottom-left corner. This...

Building an experiment

In this section, we'll show how to build an experiment from scratch using a custom dataset. With the GUI, creating new experiments is very fast and results can be viewed immediately. Azure ML Studio contains modules for all common ML and data-processing tasks, so it is a great tool for testing ideas quickly and iteratively. If the built-in modules are not sufficient for the task at hand, the script modules can be used for improved extensibility, explained as follows.

Importing and preprocessing data

As already discussed, Azure ML Studio is a complete ML tool that takes care of every step in the ML model development process. The only input needed is a raw dataset in a format understood by ML Studio...

Deploying a model as a web service

One of the biggest strengths of Azure ML Studio is the ease with which you can deploy models to the cloud, to be consumed by other applications. Once an ML model is trained, as demonstrated in the previous section, it can be exported to ML Studio Web Services with just a few clicks. Deployment creates a web API for the model, which can be called from any internet-connected application. The model takes the features as input data and produces a predicted value as output. By deploying models to the ML Studio Web Service, there is no need to worry about the underlying server infrastructure. The computing resources and maintenance are handled entirely by Azure.

The following subsections show how to deploy an already trained model to the web service and how to test a model with user input. In the final subsection, we'll show how to import and...

Summary

Azure Machine Learning Studio is a fully managed platform for developing machine learning models, enabling the user to concentrate on the essential tasks and problems in machine learning development. The graphical user interface is easy to learn and its usage requires no programming skills. Even users who have no prior experience in programming or machine learning can learn to use it, and the Experiment template collection contains many real-world examples of ML models to start with. ML Studio is a great way to start learning to develop ML models, and the sample datasets in ML Studio make it possible to develop your own models, even if you don't have your own data to start with.

Machine Learning Studio contains modules for all the most common ML-related tasks, such as data preprocessing, tuning hyperparameters, and evaluating the performance of ML algorithms. If the...