Chapter 12: Realizing Business Value with AutoML

You have acquired a wide variety of technical skills throughout this book. You're now able to train regression, classification, and forecasting models with AutoML. You can code AutoML solutions in Python using Jupyter notebooks, you know how to navigate Azure Machine Learning Studio, and you can even integrate machine learning pipelines in Azure Data Factory (ADF). Yet, technical skills alone will not guarantee the success of your projects. In order to realize business value, you have to gain the trust and acceptance of your end users. 

In this chapter, you will begin by learning how to present end-to-end architectures in a way that makes it easy for end users to understand. Then, you will learn which visualizations and metrics to use to show off your model's performance, after which you will learn how to visualize and interpret AutoML's built-in explainability function.

You will also explore options to run...

Technical requirements

In this chapter, you will use models that you created in previous chapters to retrieve graphs, charts, and metrics. As such, you will need a working internet connection, an Azure account, and an AMLS workspace. You will also need to complete the exercises in Chapter 4, Building an AutoML Regression Solution, and Chapter 5, Building an AutoML Classification Solution.

The following are the prerequisites for the chapter:

• Access to the internet.
• A web browser, preferably Google Chrome or Microsoft Edge Chromium.
• A Microsoft Azure account.
• An AMLS workspace.
• You need to have trained and registered the Diabetes-AllData-Regression-AutoML machine learning model in Chapter 4, Building an AutoML Regression Solution.
• You need to have trained and registered the Iris-Multi-Classification machine learning model in Chapter 5, Building an AutoML Classification Solution.

There is no new code for this chapter.

Architecting AutoML solutions

Architecting AutoML solutions refers to drawing end-to-end diagrams. These act as blueprints for how you should build out your solution, and also can be used to explain to your end users how everything works. While many IT solutions are complex and can take many forms, AutoML-based solutions follow standard patterns that require you to make a few important decisions.

In this section, you'll first learn what decisions to make before architecting a decision. Then, you will learn how to architect an end-to-end batch scoring solution and an end-to-end real-time scoring solution that's easy to explain to end users. Although the architecture may be simplified, the more standard it is, the easier it is to implement, explain, and understand.

Making key architectural decisions for AutoML solutions

When drawing an architectural diagram, there are several key considerations you need to make, the most important being whether you need to make a batch...

Visualizing AutoML modeling results

Presenting the results of your AutoML model to your business is integral to the adoption of your solution. After all, your end users are unlikely to adopt your solution unless they can be sure that it meets certain standards of performance. There are many ways of presenting the results of ML models; the most effective way of presenting your results is through visualizations.

Thankfully, AutoML runs provide automatic visualizations for results of regression, classification, and forecasting. Regression and forecasting share identical visualizations, while classification is quite different. In each case, you only want to share a single visualization with your end user; multiple views of the same results are likely to only cause confusion.

In this section, you'll first uncover what to show your end user for classification before moving onto regression and forecasting.

Visualizing the results of classification

Confusion matrices, as...

Explaining AutoML results to your business

To realize business value, your AutoML models must be implemented and used by the business. A common obstacle to implementation is a lack of trust stemming from a lack of understanding of how ML works. At the same time, explaining the ins and outs of how individual ML algorithms work is a poor way to gain trust. Throwing math symbols and complicated statistics at end users will not work unless they already have a deep background in mathematics.

Instead, use AutoML's inbuilt explainability. As long as you enable explainability when training models, you can say exactly which features AutoML is using to generate predictions. In general, it's a good practice to do the following four things:

• Always enable explainability when training any AutoML model.
• When presenting results to the business, first show performance, then show explainability.
• Rank the features in order of most to least important.
• Drop any unimportant...

Using AutoML in other Microsoft products

In this book, you've learned how to use AutoML on Azure, but you can also use AutoML in a wider suite of Microsoft products. While you can easily create and productionalize just about any AutoML solution following the architectural patterns in the Architecting AutoML solutions section of this chapter, there are certain scenarios in which you may want to use AutoML on other Microsoft platforms. You can find AutoML in the following places:

• PowerBI
• Azure Synapse Analytics
• ML.NET
• HDInsight
• SQL Server
• Azure Databricks

Even though AutoML is available for these services, there are many differences of which you should be aware. Some services are code-free while others are code-only. Some services preclude you from training forecasting algorithms and others are based on entirely different ML frameworks. In this section, you will be guided through the general capabilities service by service.

Using AutoML within...

Realizing business value

Realizing business value ultimately comes down to whether your business partners choose to act on the predictions of your ML models. Without action, the work of data scientists amounts to little more than a science experiment. Your business partners must be motivated and willing to make your predictions a part of their decision-making process. Gaining their trust is paramount.

In order to gain the trust of your company's decision-making leadership, you first have to ascertain what kind of solution you are building with AutoML. Some solutions are rather easily and rapidly adopted, while others are likely to encounter hard resistance.

There are key two factors that determine how readily your AutoML solution is accepted: whether your tool is replacing an existing solution and whether your tool is directly involved in an automated decision process or is assisting human decision makers. Figure 12.8 shows how difficult it is to gain acceptance based...

Summary

Gaining end user acceptance can be difficult, but having the right approach can make it a lot easier. Walking your end users through an architectural diagram, carefully explaining the model's performance to them using the right metrics, and spending time explaining which features the model is using to make predictions are all key to selling your solution to end users. Furthermore, you can tailor your message based on what type of solution you are building to gain end user trust.

You are now at the end of the book and I'd like you to reflect on the journey. You've acquired many technical skills, including the ability to train AutoML models, deploy AutoML models for scoring in batch and real-time scoring, and design, create, and implement full end-to-end AutoML solutions. You also have an approach to sell those solutions to your business partners, gain their trust, and, ultimately, realize value. By crafting powerful solutions with AutoML on Azure, you&apos...

Why subscribe?

• Spend less time learning and more time coding with practical eBooks and Videos from over 4,000 industry professionals
• Improve your learning with Skill Plans built especially for you
• Get a free eBook or video every month
• Fully searchable for easy access to vital information
• Copy and paste, print, and bookmark content

Did you know that Packt offers eBook versions of every book published, with PDF and ePub files available? You can upgrade to the eBook version at packt.com and as a print book customer, you are entitled to a discount on the eBook copy. Get in touch with us at [e-mail] for more details.

At www.packt.com, you can also read a collection of free technical articles, sign up for a range of free newsletters, and receive exclusive discounts and offers on Packt books and eBooks.