Loading the video player ...
If this message stays on, please ensure JavaScript is enabled and that you are using a supported device and browser. Ad blockers may also cause this issue.
For more information please visit our device support page.
Microsoft Azure Machine Learning, or just Azure ML, is a cloud-based platform for designing, developing, testing and deploying predictive models. Those models can use many well-known machine learning algorithms, such as decision forests or neural networks, to find useful patterns in your data. You can also use your own or third-party code written in the powerful and modern statistical analysis language simply known as R, or run your Python code, especially if you like using its popular data science packages, such as pandas, numpy, scipy or scikit-learn (part of the Anaconda distribution).
The heart of Azure ML toolkit is ML Studio, a web-based graphical environment where you upload or connect to data, and where you can create and validate your models, which Azure ML refers to as experiments. This is an apt name, because an experiment can consist of multiple models and much additional logic, or other helpful calculations and data preparation tasks or output processing steps. Our video shows a simple experiment (2:09) which is used to find a pattern that determines if a prospective customer is likely to make multiple purchases from our fictitious car dealership store, based on the data representing customer lifetime purchases and lifetime value (this data set is available to all Full Access Members to download from here).
Once you have a model that works and which you have successfully validated (validity has been discussed in the previous module), you may want to deploy it into production! The 21st century way of doing that is, of course, with the help of the cloud, and this is where Microsoft Azure ML comes into its forte: with a few clicks you can turn an experiment into an externally callable, high-performance web service. You will briefly see how to call and test this service in this video (5:44). Our service can predict if a customer is likely to make multiple purchases, and so it could be used as part of a web application, perhaps to decide if to offer this customer a particularly tasty promotion or maybe to direct them to a special page describing the values of our top-shelf products. Nevertheless, if you want to learn how to do it properly, you ought to follow the next, in-depth module, which explains why it is better to create a separate Azure ML experiment, one which is used with a previously trained and validated model, as the base for your production web services.
If you would like to read a bit more about Azure ML, make sure to have a look at this essay, which discusses the Azure ML concepts in more detail. Above, all, make sure to follow the remaining modules in this course on data science, starting with the fundamental concepts, such as cases, and algorithms, if you are new to this fascinating discipline. The best way to take advantage of this and much of the upcoming content is to be one of our Full Access Members.
Comments
Iman Ahmadi · 19 December 2014
Rafal, as always thank you so much for your hard work and great videos.
Rafal Lukawiecki · 19 December 2014
Many thanks, Iman. Always a pleasure to hear from you.
joseusul · 18 February 2015
Hi Rafa,
Congratulations, a great explanation again.
One question. I have created a free trial Azure ML account during 30 days.
In order to test this technology, Can you advise us what's the best way to upgrade this limited account? I mean, I am interested in testing some experiments with limited resources, not really big data projects for now.
Thanks in advance for your answer.
José.
Rafal Lukawiecki · 18 February 2015
José, I would suggest you go for the "paid" Azure account, that is, associate a credit card with it, and pay as you go. If you only do testing and small-scale development you will find the charges are very low, single dollars per month, and there is no recurring subscription fee.
joseusul · 20 February 2015
Thank you very much Rafa.
Kind regards.