Azure Machine Learning, Study Guides, Projects, Research of Electrical Circuit Analysis

Download Azure Machine Learning and more Study Guides, Projects, Research Electrical Circuit Analysis in PDF only on Docsity!

Azure Machine

Learning

Microsoft Azure Essentials

Jeff Barnes

- Hundreds of titles available – Books, eBooks, and online

resources from industry experts

**- Free U.S. shipping

• eBooks in multiple formats** – Read on your computer,

tablet, mobile device, or e-reader

**- Print & eBook Best Value Packs

• eBook Deal of the Week** – Save up to 60% on featured titles - Newsletter and special offers – Be the first to

hear about new releases, specials, and more

- Register your book – Get additional benefits

microsoftpressstore.com

Visit us today at

Wait, there’s more...

Find more great content and resources in the

Microsoft Press Guided Tours app.

The Microsoft Press Guided Tours app provides

insightful tours by Microsoft Press authors of new and

evolving Microsoft technologies.

Download from

Windows Store

Download from

Windows Store

• Share text, code, illustrations, videos, and links with

peers and friends

• Create and manage highlights and notes
• View resources and download code samples
• Tag resources as favorites or to read later
• Watch explanatory videos
• Copy complete code listings and scripts

PUBLISHED BY Microsoft Press A division of Microsoft Corporation One Microsoft Way Redmond, Washington 98052-

Copyright © 2015 Microsoft Corporation. All rights reserved.

No part of the contents of this book may be reproduced or transmitted in any form or by any means without the written permission of the publisher.

ISBN: 978-0-7356-9817-

Microsoft Press books are available through booksellers and distributors worldwide. If you need support related to this book, email Microsoft Press Support at [email protected]. Please tell us what you think of this book at http://aka.ms/tellpress.

This book is provided “as-is” and expresses the authors’ views and opinions. The views, opinions, and information expressed in this book, including URL and other Internet website references, may change without notice.

Unless otherwise noted, the companies, organizations, products, domain names, e-mail addresses, logos, people, places, and events depicted in examples herein are fictitious. No association with any real company, organization, product, domain name, e-mail address, logo, person, place, or event is intended or should be inferred.

Microsoft and the trademarks listed at http://www.microsoft.com on the “Trademarks” webpage are trademarks of the Microsoft group of companies. All other marks are property of their respective owners.

Acquisitions, Developmental, and Project Editor: Devon Musgrave Editorial Production: nSight, Inc. Copyeditor: Teresa Horton Cover: Twist Creative

Foreword

I’m thrilled to be able to share these Microsoft Azure Essentials ebooks with you. The power that Microsoft Azure gives you is thrilling but not unheard of from Microsoft. Many don’t realize that Microsoft has been building and managing datacenters for over 25 years. Today, the company’s cloud datacenters provide the core infrastructure and foundational technologies for its 200-plus online services, including Bing, MSN, Office 365, Xbox Live, Skype, OneDrive, and, of course, Microsoft Azure. The infrastructure is comprised of many hundreds of thousands of servers, content distribution networks, edge computing nodes, and fiber optic networks. Azure is built and managed by a team of experts working 24x7x365 to support services for millions of customers’ businesses and living and working all over the globe.

Today, Azure is available in 141 countries, including China, and supports 10 languages and 19 currencies, all backed by Microsoft's $15 billion investment in global datacenter infrastructure. Azure is continuously investing in the latest infrastructure technologies, with a focus on high reliability, operational excellence, cost-effectiveness, environmental sustainability, and a trustworthy online experience for customers and partners worldwide.

Microsoft Azure brings so many services to your fingertips in a reliable, secure, and environmentally sustainable way. You can do immense things with Azure, such as create a single VM with 32TB of storage driving more than 50,000 IOPS or utilize hundreds of thousands of CPU cores to solve your most difficult computational problems.

Perhaps you need to turn workloads on and off, or perhaps your company is growing fast! Some companies have workloads with unpredictable bursting, while others know when they are about to receive an influx of traffic. You pay only for what you use, and Azure is designed to work with common cloud computing patterns.

From Windows to Linux, SQL to NoSQL, Traffic Management to Virtual Networks, Cloud Services to Web Sites and beyond, we have so much to share with you in the coming months and years.

I hope you enjoy this Microsoft Azure Essentials series from Microsoft Press. The first three ebooks cover fundamentals of Azure, Azure Automation, and Azure Machine Learning. And I hope you enjoy living and working with Microsoft Azure as much as we do.

Introduction

Microsoft Azure Machine Learning (ML) is a service that a developer can use to build predictive analytics models (using training datasets from a variety of data sources) and then easily deploy those models for consumption as cloud web services. Azure ML Studio provides rich functionality to support many end-to-end workflow scenarios for constructing predictive models, from easy access to common data sources, rich data exploration and visualization tools, application of popular ML algorithms, and powerful model evaluation, experimentation, and web publication tooling.

This ebook will present an overview of modern data science theory and principles, the associated workflow, and then cover some of the more common machine learning algorithms in use today. We will build a variety of predictive analytics models using real world data, evaluate several different machine learning algorithms and modeling strategies, and then deploy the finished models as machine learning web service on Azure within a matter of minutes. The book will also expand on a working Azure Machine Learning predictive model example to explore the types of client and server applications you can create to consume Azure Machine Learning web services.

The scenarios and end-to-end examples in this book are intended to provide sufficient information for you to quickly begin leveraging the capabilities of Azure ML Studio and then easily extend the sample scenarios to create your own powerful predictive analytic experiments. The book wraps up by providing details on how to apply “continuous learning” techniques to programmatically “retrain” Azure ML predictive models without any human intervention.

Who should read this book

This book focuses on providing essential information about the theory and application of data science principles and techniques and their applications within the context of Azure Machine Learning Studio. The book is targeted towards both data science hobbyists and veterans, along with developers and IT professionals who are new to machine learning and cloud computing. Azure ML makes it just as approachable for a novice as a seasoned data scientist, helping you quickly be productive and on your way towards creating and testing machine learning solutions.

Detailed, step-by-step examples and demonstrations are included to help the reader understand how to get started with each of the key predictive analytic algorithms in use today and their corresponding implementations in Azure ML Studio. This material is useful not only for those who have no prior experience with Azure Machine Learning, but also for those who are experienced in the field of data science. In all cases, the end-to-end demos help reinforce the machine learning concepts with concrete examples and real-life scenarios. The chapters do build on each other to some extent; however, there is no requirement that you perform the hands-on demonstrations from previous

 Chapter 5, “Regression analytics,” takes a deeper look at some of the more advanced machine learning algorithms that are exposed in Azure ML Studio.

 Chapter 6, “Cluster analytics,” explores scenarios where the machine conducts its own analysis on the dataset, determines relationships, infers logical groupings, and generally attempts to make sense of chaos by literally determining the forests from the trees.

 Chapter 7, “The Azure ML Matchbox recommender,” explains one of the most powerful and pervasive implementations of predictive analytics in use today on the web today and how it is crucial to success in many consumer industries.

 Chapter 8, “Retraining Azure ML models,” explores the mechanisms for incorporating “continuous learning” into the workflow for our predictive models.

Conventions and features in this book

This book presents information using the following conventions designed to make the information readable and easy to follow:

 To create specific Azure resources, follow the numbered steps listing each action you must take to complete the exercise.

 There are currently two management portals for Azure: the Azure Management Portal at http://manage.windowsazure.com and the new Azure Preview Portal at http://portal.azure.com. This book assumes the use of the original Azure Management Portal in all cases.

 A plus sign (+) between two key names means that you must press those keys at the same time. For example, “Press Alt+Tab” means that you hold down the Alt key while you press Tab.

System requirements

For many of the examples in this book, you need only Internet access and a browser (Internet Explorer 10 or higher) to access the Azure portal. Chapter 4, “Creating Azure ML client and server applications,” and many of the remaining chapters use Visual Studio to show client applications and concepts used in developing applications for consuming Azure Machine Learning web services. For these examples, you will need Visual Studio 2013. You can download a free copy of Visual Studio Express at the link below. Be sure to scroll down the page to the link for “Express 2013 for Windows Desktop”: http://www.visualstudio.com/en-us/products/visual-studio-express-vs.aspx

The following are system requirements:

 Windows 7 Service Pack 1, Windows 8, Windows 8.1, Windows Server 2008 R2 SP1, Windows

Server 2012, or Windows Server 2012 R

 Computer that has a 1.6GHz or faster processor (2GHz recommended)

 1 GB (32 Bit) or 2 GB (64 Bit) RAM (Add 512 MB if running in a virtual machine)

 20 GB of available hard disk space

 5400 RPM hard disk drive

 DirectX 9 capable video card running at 1024 x 768 or higher-resolution display

 DVD-ROM drive (if installing Visual Studio from DVD)

 Internet connection

Depending on your Windows configuration, you might require Local Administrator rights to install or configure Visual Studio 2013.

Acknowledgments

This book is dedicated to my father who passed away during the time this book was being written, yet wisely predicted that computers would be a big deal one day and that I should start to “ride the wave” of this exciting new field. It has truly been quite a ride so far.

This book is the culmination of many long, sacrificed nights and weekends. I’d also like to thank my wife Susan, who can somehow always predict my next move long before I make it. And to my children, Ryan, Brooke, and Nicholas, for their constant support and encouragement.

Special thanks to the entire team at Microsoft Press for their awesome support and guidance on this journey. Most of all, it was a supreme pleasure to work with my editor, Devon Musgrave, who provided constant advice, guidance, and wisdom from the early days when this book was just an idea, all the way through to the final copy. Brian Blanchard was also critical to the success of this book as his keen editing and linguistic magic helped shape many sections of this book.

Errata, updates, & support

We’ve made every effort to ensure the accuracy of this book. You can access updates to this book—in the form of a list of submitted errata and their related corrections—at:

http://aka.ms/AzureML/errata

If you discover an error that is not already listed, please submit it to us at the same page.

Stay in touch

Let’s keep the conversation going! We’re on Twitter: http://twitter.com/MicrosoftPress

Chapter 1

Introduction to the science of data

Welcome to the exciting new world of Microsoft Azure Machine Learning! Whether you are an expert data scientist or aspiring novice, Microsoft has unleashed a powerful new set of cloud-based tools to allow you to quickly create, share, test, train, fail, fix, retrain, and deploy powerful machine learning experiments in the form of easily consumable Web services, all built with the latest algorithms for predictive analytics. From there, you can fine-tune your experiments by continuously “training” them with new data sets for maximum results.

Bill Gates once said, “A breakthrough in machine learning would be worth ten Microsofts,” and the new Azure Machine Learning service takes on that ambitious challenge with a truly differentiated cloud-based offering that allows easy access to the tools and processing workflow that today’s data scientist needs to be quickly successful. Armed with only a strong hypothesis, a few large data sets, a valid credit card, and a browser, today’s machine learning entrepreneurs are learning how to mine for gold inside many of today’s big data warehouses.

What is machine learning?

Machine learning can be described as computing systems that improve with experience. It can also be described as a method of turning data into software. Whatever term is used, the results remain the same; data scientists have successfully developed methods of creating software “models” that are trained from huge volumes of data and then used to predict certain patterns, trends, and outcomes.

Predictive analytics is the underlying technology behind Azure Machine Learning, and it can be simply defined as a way to scientifically use the past to predict the future to help drive desired outcomes.

Machine learning and predictive analytics are typically best used under certain circumstances, as they are able to go far beyond standard rules engines or programmatic logic developed by mere mortals. Machine learning is best leveraged as means to optimize a desired output or prediction using example or past historical experiential data. One of the best ways to describe machine learning is to compare it with today’s modern computer programming paradigms.

Under traditional programming models, programs and data are processed by the computer to produce a desired output, such as using programs to process data and produce a report (see Figure 1-1).

pack. By combining these deep psychological motivators with the right historical transaction data and then applying optimized filtering algorithms, you can easily see how to implement a highly effective e-commerce up-sell strategy.

One of humankind’s most basic and powerful natural instincts is the fear of missing out on something, especially if others are doing it. This is the underlying foundation of social networks, and nowhere is predictive analytics more useful and effective than in helping to predict human nature in conjunction with the Web. By combining this deep, innate psychological desire with the right historical transaction data and then applying optimized filtering algorithms, you can implement a highly effective e-commerce upselling strategy.

Let’s think about the underlying data requirements for this highly effective prediction algorithm to work. The most basic requirement is a history of previous orders, so the system can check for other items that were bought together with the item the user is currently viewing. By then combining and filtering that basic data (order history) with additional data attributes from a user’s profile like age, sex, marital status, and zip code, you can create a more deeply targeted set of recommendations for the user.

But wait, there’s more! What if you could have also inferred the user’s preferences and buying patterns based on the category and subcategory of items he or she has bought in the past? Someone who purchases a bow, arrows, and camping stove can be assumed to be a hunter, who most likely also likes the outdoors and all that entails, like camping equipment, pick-up trucks, and even marshmallows.

This pattern of using cojoined data to infer additional data attributes is where the science of data really takes off, and it has serious financial benefits to organizations that know how to leverage this technology effectively. This is where data scientists can add the most value, by aiding the machine learning process with valuable data insights and inferences that are (still) more easily understood by humans than computers.

This is also where it becomes most critical to have the ability to rapidly test a hunch or theory to either “fail-fast” or confirm the logic of your prediction algorithms, and really fine-tune a prediction model. Fortunately, this is an area in which Azure Machine Learning really shines. In later chapters, we will learn about how you can quickly create, share, deploy, and test Azure Machine Learning experiments to rapidly deploy predictive analytics in your organization.

In a way, Azure Machine Learning could be easily compared with training children or animals, without the need for food, water, or physical rest, of course. Continuous and adaptive improvement is one of the primary hallmarks of the theory of evolution and Darwinism; in this case, it represents a major milestone in the progression of computational theory and machine learning capabilities.

Machine learning could then be compared to many of the concepts behind evolution itself; specifically how, given enough time and data (in the form of real-world experiences), organisms in the natural world can overcome changes in the environment through genetic and behavioral adaptations. The laws of nature have always favored the notion of adaptation to maximize the chances of survival.

Today’s perfect storm for machine learning

Today’s modern predictive analytics systems are achieving this same level of machine evolution much more rapidly due to the following industry trends:

 Exponential data growth

 We are virtually sitting on mountains of highly valuable historical transactional data, most of it digitally archived and readily accessible.

 There is an increasing abundance of real-time data via embedded systems and the evolution of “the Internet of Things” (IoT) connected devices.

 We have an ability to create new synthetic data via extrapolation and projection of existing historical data to create realistic simulated data.

 Cheap global digital storage

 Vast quantities of free or low-cost, globally available, digital storage are readily accessible over the Web today.

 From personal devices to private and public clouds, we have access to multiple storage mechanisms to house all our never-ending streams of data.

 Ubiquitous computing power

 Cloud computing services are everywhere today and readily available through a large selection of cloud and hosting partners, all at competitive rates.

 Access is simple. A credit card and a browser are all you need to get started and pay by the hour or minute for everything you need to get started.

 The rise of big data analytics

 The economic powers of predictive analytics in many real-world business-use cases, many with extremely favorable financial outcomes, are being realized.

To that end, one of the most intriguing aspects of machine learning is that it is always adaptive and always learning from any mistakes or miscalculations. As a result, a good feedback/correction loop is essential for fine-tuning a predictive model. The advent of cheap cloud storage and ever increasingly ubiquitous computing power make it easier to quickly and efficiently mine for gold in your data.