Preface

It is not so often in life that you get a second chance. I remember that only days after we stopped editing the first edition, I kept asking myself, "Why didn't I...?", or "What the heck was I thinking saying it like that?", and on and on. In fact, the first project I started working on after it was published had nothing to do with any of the methods in the first edition. I made a mental note that if given the chance, it would go into a second edition.

When I started with the first edition, my goal was to create something different, maybe even create a work that was a pleasure to read, given the constraints of the topic. After all the feedback I received, I think I hit the mark. However, there is always room for improvement, and if you try and be everything to all people, you become nothing to everybody. I'm reminded of one of my favorite Frederick the great quotes, "He who defends everything, defends nothing". So, I've tried to provide enough of the skills and tools, but not all of them, to get a reader up and running with R and machine learning as quickly and painlessly as possible. I think I've added some interesting new techniques that build on what was in the first edition. There will probably always be the detractors who complain it does not offer enough math or does not do this, that, or the other thing, but my answer to that is they already exist! Why duplicate what was already done, and very well, for that matter? Again, I have sought to provide something different, something that would keep the reader's attention and allow them to succeed in this competitive field.

Before I provide a list of the changes/improvements incorporated into the second edition, chapter by chapter, let me explain some universal changes. First of all, I have surrendered in my effort to fight the usage of the assignment operator <- versus just using =. As I shared more and more code with others, I realized I was out on my own using = and not <-. The first thing I did when under contract for the second edition was go line by line in the code and change it. The more important part, perhaps, was to clean and standardize the code. This is also important when you have to share code with coworkers and, dare I say, regulators. Using RStudio facilitates this standardization in the most recent versions. What sort of standards! Well, the first thing is to properly space the code. For instance, I would not hesitate in the past to write c(1,2,3,4,5,6). Not anymore! Now, I will write this--c(1, 2, 3, 4, 5, 6)--as a space after commas, which makes it easier to read. If you want other ideas, please have a look a Google's R style guide, https://google.github.io/styleguide/Rguide.xml/. I also received a number of e-mails saying that the data I scraped off the Web wasn't available. The National Hockey League decided to launch a completely new version of their statistics, so I had to start from scratch. Problems such as that led me to put data on GitHub.

All in all, I put forth a rather large effort to put the best possible tool in your hands to get you going. On another note, in the month of February '17, there was much attention on the Web on these comments from entrepreneur Mark Cuban:

• "Artificial Intelligence, deep learning, machine learning--whatever you’re doing if you don’t understand it--learn it. Because otherwise you’re going to be a dinosaur within 3 years."
• "I personally think there's going to be a greater demand in 10 years for liberal arts majors than there were for programming majors and maybe even engineering, because when the data is all being spit out for you, options are being spit out for you, you need a different perspective in order to have a different view of the data. And so is having someone who is more of a freer thinker."

Besides the fact that these comments created a bit of a stir on the blogosphere, they also seem to be, at first glance, mutually exclusive. But think about what he is saying here. I think he gets to the core of why I felt compelled to write this book. Here is what I believe, machine learning needs to be embraced and utilized, to some extent, by the masses: the tired, the poor, the hungry, the proletariat, and the bourgeoisie. More and more availability of computational power and information will make machine learning something for virtually everyone. However, the flip side of that and what, in my mind, has been and will continue to be a problem is the communication of results. What are you going to do when you describe true positive rate and false positive rate and receive blank stares? How do you quickly tell a story that enlightens your audience? If you think it can't happen, please drop me a note, I'd be more than happy to share my story.

We must have people who can lead these efforts and influence their organization. If a degree in history or music appreciation helps in that endeavor, then so be it. I study history every day, and it has helped me tremendously. Cuban's comments have reinforced my belief that in many ways, the first chapter is the most important in this book. If you are not asking your business partners "what they plan to do differently", you'd better start tomorrow. There are far too many people working far too hard to complete an analysis that is completely irrelevant to the organization and its decisions.