John D. Kelleher

I have generally enjoyed books in the MIT Press Essential Knowledge Series, but this title is the weakest of those that I have read so far. Other titles in the series did a good job of summarizing the field of study. This one felt like it only barely scratched the surface, and provided examples that were far too simple and obvious.

It also made me question why this field is called "Data Science". The book doesn't really demonstrate how this discipline is a branch of science by any definition of that term (see, for example, Lee McIntyre's The Scientific Attitude for an exploration of what science is).… (plus d'informations)

I genuinely applaud the authors for this book. Their work is highly commendable.

For a novice like me, even reading the appendices was helpful. These fundamentals are definitely key to understanding the many advanced topics throughout the book. The authors disclose other essential concepts like co-variance and correlation in the chapters as well. This mandates reading each section of the book carefully and thoroughly.

The book showcases illustrious work on CRISP-DM methodology, detailing the nuances of each phase. The authors walk through each of these five phases using real-world examples. The case studies towards the end of the book revise major portions of the book.

For example, in R language, we can just call the 'lm' function to know the intercept and the weights. This book does not abstract that from us but instead focuses on the actual implementation of the algorithms and the mathematics involved behind getting the answer. There are quite a few articles/blogs which tell us how to implement an algorithm using programming languages. But knowing the inner-working is always better.

The only question I have for the authors is why H-matrix was not introduced to find the weights in the linear regression.

I absolutely enjoyed reading this book and learned quite a lot about machine learning.… (plus d'informations)