Azure and Excel - Machine Learning in the Cloud

The clear tendency nowadays is to move all analysis, storage, and visualization activities to the cloud. In this chapter, you will find information about how to use Azure services and get a free subscription to test them. Deep learning seems to be the path to general Artificial Intelligence, that is, machines that can think as humans. We are not even close to that yet, but artificial neural networks are used in computer vision, text and speech analysis, and many other advanced applications. There are many use cases for artificial intelligence that are pre-built inside Azure and that can be used by building an experiment, as shown in detail in this chapter.

The following topics will be covered in this chapter:

• Introducing the Azure Cloud
• Using Azure Machine Learning Studio for free – a step-by-step guide
• Loading your data...

Technical requirements

To complete this chapter, you will need a Microsoft account. If you don't have one, you can get one for free at https://signup.live.com/.

You will also need to download the titanic_small.csv file from the GitHub repository at https://github.com/PacktPublishing/Hands-On-Machine-Learning-with-Microsoft-Excel-2019/tree/master/Chapter10.

Introducing the Azure Cloud

Cloud computing is the on demand availability of computer system resources, especially data storage and computing power, without direct active management by the user. The term is generally used to describe data centers available to many users over the Internet. Here are some advantages of using the cloud instead of on-premises computers:

• Cost: Instead of buying and maintaining expensive hardware and software, the cloud model offers to pay only for what is used.
• Speed: A large amount of resources can be obtained typically in minutes, just by configuring some settings in a website.
• Global scale: The size and location of the resources can be dynamically changed according to the user's needs.
• Productivity: IT staff can save time and focus on tasks that help the business grow, instead of doing hands-on maintenance of local equipment.
• Performance: The...

Using AMLS for free – a step-by-step guide

Microsoft AMLS is a tool that provides a drag and drop interface to build, test, and deploy machine learning models and analytics solutions. It is possible to publish models as web services that can be consumed from Excel (among other tools).

We will start by registering in the AMLS home page, using your Microsoft account:

1. Open https://studio.azureml.net/. You will see the following front page:

2. Click on Sign up here and you will get to the next page:

3. Choose the second option (that is, Free Workspace), which requires that you have a Microsoft account. The advantage is that this option is free and you can use it without limits. Once you click on the Free Workspace option, you will be taken to the following login screen:

4. Enter the username you chose when you created your Microsoft account and click on Next. You will then...

Loading your data into AMLS

There is no machine learning project without data, so the first step in our analysis is to load the input file (titanic_small.csv) into AMLS. This is a simplified version of the Titanic dataset, which contains three features and one target variable:

• Features:
  • pclass: The class in which the passenger traveled (values 1, 2, or 3 corresponding to 1st, 2nd, and 3rd class)
  • sex: Passenger's gender (female or male)
  • Age group: Infant, child, teenager, adult, elderly, or unknown
• Target variable:
  • Survived: 1 if the passenger survived the shipwreck, 0 if they didn't.

To load the file, follow these steps:

1. From the home page, click on DATASETS. You will see an empty list of datasets:

2. Click on +NEW to get a link to upload a local data file:

3. Click on FROM LOCAL FILE and you will see the following dialog box:

4. Click on Choose File and navigate...

Creating and running an experiment in AMLS

The basic components of AMLS are experiments. They are built by dragging and dropping predefined modules into a workspace. Each module has some defined task, a given number of parameters that can be chosen at runtime, and a defined number of input and output nodes. Here is a screenshot of the AMLS module:

They can be connected to build an analysis workflow, from data input and transformation to machine learning model training and results. We will go step by step and create a machine learning experiment, training a decision tree to predict the survival of the Titanic passengers.

Creating a new experiment

Perform the following steps to create a new environment:

1. From the home page...

Summary

We went through all the necessary steps to open an account in AMLS, a part of the Microsoft Azure cloud that helps us to build simple data and analysis flows. We also built two experiments: one of them trains a decision tree and the other predicts the target variable. Then we learned how to create a web service and connect Excel to it, sending and receiving data.

In the next chapter, we will show the present status of machine learning, which almost completely moves operations to the cloud, makes the data flows completely automatic, and uses automation to fine-tune the predictive models.

Questions

1. What are the main advantages of using cloud computing?
2. Is cloud computing only useful for machine learning?
3. What is a web service and why is it useful?
4. If the model is already trained, why do we need to include the input data model in the analysis flow used for prediction?
5. Why did we split training and prediction into two different flows?

Further reading

Check out the following resources for more information on the topics covered in this chapter:

• Understanding the Basics of Cloud Computing: https://www.lucidchart.com/blog/cloud-computing-basics
• Azure Machine Learning Studio Documentation: https://docs.microsoft.com/en-us/azure/machine-learning/studio/
• Decision Trees Explained Easily: https://medium.com/@chiragsehra42/decision-trees-explained-easily-28f23241248