Mastering Machine Learning with Spark 2.x Create scalable machine learning applications to power a modern data-driven business using Spark Alex Tellez Max Pumperla Michal Malohlava BIRMINGHAM - MUMBAI Mastering Machine Learning with Spark 2.x Copyright © 2017 Packt Publishing All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, without the prior written permission of the publisher, except in the case of brief quotations embedded in critical articles or reviews. Every effort has been made in the preparation of this book to ensure the accuracy of the information presented. However, the information contained in this book is sold without warranty, either express or implied. Neither the authors, nor Packt Publishing, and its dealers and distributors will be held liable for any damages caused or alleged to be caused directly or indirectly by this book. Packt Publishing has endeavored to provide trademark information about all of the companies and products mentioned in this book by the appropriate use of capitals. However, Packt Publishing cannot guarantee the accuracy of this information. First published: August 2017 Production reference: 1290817 Published by Packt Publishing Ltd. Livery Place 35 Livery Street Birmingham B3 2PB, UK. ISBN 978-1-78528-345-1 www.packtpub.com Credits Author Copy Editor Alex Tellez Muktikant Garimella Max Pumperla Michal Malohlava Reviewer Project Coordinator Dipanjan Deb Ulhas Kambali Commissioning Editor Proofreader Veena Pagare Safis Editing Acquisition Editor Indexer Larissa Pinto Rekha Nair Content Development Editor Graphics Nikhil Borkar Jason Monteiro Technical Editor Production Coordinator Diwakar Shukla Melwyn Dsa About the Authors Alex Tellez is a life-long data hacker/enthusiast with a passion for data science and its application to business problems. He has a wealth of experience working across multiple industries, including banking, health care, online dating, human resources, and online gaming. Alex has also given multiple talks at various AI/machine learning conferences, in addition to lectures at universities about neural networks. When he’s not neck-deep in a textbook, Alex enjoys spending time with family, riding bikes, and utilizing machine learning to feed his French wine curiosity! First and foremost, I’d like to thank my co-author, Michal, for helping me write this book. As fellow ML enthusiasts, cyclists, runners, and fathers, we both developed a deeper understanding of each other through this endeavor, which has taken well over one year to create. Simply put, this book would not have been possible without Michal’s support and encouragement. Next, I’d like to thank my mom, dad, and elder brother, Andres, who have been there every step of the way from day 1 until now. Without question, my elder brother continues to be my hero and is someone that I will forever look up to as being a guiding light. Of course, no acknowledgements would be finished without giving thanks to my beautiful wife, Denise, and daughter, Miya, who have provided the love and support to continue the writing of this book during nights and weekends. I cannot emphasize enough how much you both mean to me and how you guys are the inspiration and motivation that keeps this engine running. To my daughter, Miya, my hope is that you can pick this book up and one day realize that your old man isn’t quite as silly as I appear to let on. Last but not least, I’d also like to give thanks to you, the reader, for your interest in this exciting field using this incredible technology. Whether you are a seasoned ML expert, or a newcomer to the field looking to gain a foothold, you have come to the right book and my hope is that you get as much out of this as Michal and I did in writing this work. Max Pumperla is a data scientist and engineer specializing in deep learning and its applications. He currently works as a deep learning engineer at Skymind and is a co- founder of aetros.com. Max is the author and maintainer of several Python packages, including elephas, a distributed deep learning library using Spark. His open source footprint includes contributions to many popular machine learning libraries, such as keras, deeplearning4j, and hyperopt. He holds a PhD in algebraic geometry from the University of Hamburg. Michal Malohlava, creator of Sparkling Water, is a geek and the developer; Java, Linux, programming languages enthusiast who has been developing software for over 10 years. He obtained his PhD from Charles University in Prague in 2012, and post doctorate from Purdue University. During his studies, he was interested in the construction of not only distributed but also embedded and real-time, component-based systems, using model-driven methods and domain-specific languages. He participated in the design and development of various systems, including SOFA and Fractal component systems and the jPapabench control system. Now, his main interest is big data computation. He participates in the development of the H2O platform for advanced big data math and computation, and its embedding into Spark engine, published as a project called Sparkling Water. I would like to thank my wife, Claire, for her love and encouragement. About the Reviewer Dipanjan Deb is an experienced analytic professional with over 17 years of cumulative experience in machine/statistical learning, data mining and predictive analytics across finance, healthcare, automotive, CPG, automotive, energy, and human resource domains. He is highly proficient in developing cutting-edge analytic solutions using open source and commercial software to integrate multiple systems in order to provide massively parallelized and large-scale optimization. www.PacktPub.com For support files and downloads related to your book, please visit www.PacktPub.com. 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 www.PacktPub.comand as a print book customer, you are entitled to a discount on the eBook copy. Get in touch with us at [email protected] for more details. At www.PacktPub.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. https:/​/​www.​packtpub.​com/​mapt Get the most in-demand software skills with Mapt. Mapt gives you full access to all Packt books and video courses, as well as industry-leading tools to help you plan your personal development and advance your career. Why subscribe? Fully searchable across every book published by Packt Copy and paste, print, and bookmark content On demand and accessible via a web browser Customer Feedback Thanks for purchasing this Packt book. At Packt, quality is at the heart of our editorial process. To help us improve, please leave us an honest review on this book's Amazon page at https:/​/​www.​amazon.​com/​dp/​1785283456. If you'd like to join our team of regular reviewers, you can e-mail us at [email protected]. We award our regular reviewers with free eBooks and videos in exchange for their valuable feedback. Help us be relentless in improving our products! Table of Contents Preface 1 Chapter 1: Introduction to Large-Scale Machine Learning and Spark 7 Data science 9 The sexiest role of the 21st century – data scientist? 10 A day in the life of a data scientist 11 Working with big data 12 The machine learning algorithm using a distributed environment 13 Splitting of data into multiple machines 14 From Hadoop MapReduce to Spark 15 What is Databricks? 16 Inside the box 17 Introducing H2O.ai 18 Design of Sparkling Water 19 What's the difference between H2O and Spark's MLlib? 20 Data munging 21 Data science - an iterative process 21 Summary 22 Chapter 2: Detecting Dark Matter - The Higgs-Boson Particle 23 Type I versus type II error 23 Finding the Higgs-Boson particle 25 The LHC and data creation 25 The theory behind the Higgs-Boson 26 Measuring for the Higgs-Boson 26 The dataset 27 Spark start and data load 28 Labeled point vector 38 Data caching 38 Creating a training and testing set 40 What about cross-validation? 41 Our first model – decision tree 43 Gini versus Entropy 44 Next model – tree ensembles 51 Random forest model 52 Grid search 54 Gradient boosting machine 56