Advanced Analytics with Spark: Patterns for Learning from Data at Scale 1st Edition

TL;DR If you are looking for a intro to data science, data analysis and machine learning at scale - this is the right book. Sure, there are others, maybe more popular books from O'Reilly considering these topics, but the authors of those are using R and Python and the books are not focused on the performance and scalability. For closer details regarding Spark you can also take a look at this introductory Spark book - Learning Spark.This book presents 9 case studies of data analysis applications in various domains. The topics are diverse and the authors always use real world datasets. Beside learning Spark and a data science you will also have the opportunity to gain insight about topics like taxi traffic in NYC, deforestation or neuroscience. Without any previous exposure or contact with machine learning readers might struggle to understand certain chapters, so I think it's good idea to actually try those examples yourself while reading and Google for further details about the used methods. Many of the chapters end only with basic models, which barely outperform the baselines, so if you want to, there is a lot of space for their improvement and further work.Spark itself provides it's users with APIs in three languages - Java, Scala and Python. This books successfully covers each one of these, although you can feel slight preference of a Scala throughout the book. For Scala starters - they always explain some of the special constructs or syntax features which is in fact a nice thing. Introduction and Appendix chapters provides basic information about the Spark core, RDDs (Resilient distributed datasets) or options of running Spark - whether in cluster (Mesos, YARN, Spark's own) or standalone settings. Throughout the book you can find some really worthy tips about Spark or data analysis - like using other serializer than the Java's default (they recommend kryo), overview of data cleansing and whole machine learning pipeline. To sum up, I recommend this book to every data scientist - because it demonstrates advanced topics like workload distribution and scaling on an enjoyable examples.

This book is a good overview of potential uses of Spark, introducing different features through a sequence of vignettes. That said, it does not go in-depth into any particular aspect of Spark.The vignettes introduce a variety of topics that Spark can tackle: recommendations, graph analysis, Monte Carlo methods, by analyzing some publicly-available dataset. The analysis conducted is explained well and very useful as an introduction to the techniques they used. As an overview of the capabilities of Spark, this method excels. In addition, all code is available in the author's Github, though there are some discrepancies between the code in the repo and the book (beyond what's expected for comparing a static book and a changing git repo). This is useful for following along and replicating the analysis as well as altering their techniques to explore the data further.On the other hand, this is not an in-depth introduction to Spark as a whole. There is an appendix introducing some Spark basics, but you'll get much further with Spark's own documentation, or the other O'Reilly book, Learning Spark. Without this aspect, it becomes harder to generalize these analyses for your own purposes.

This book fills an important gap in large scale data science.Spark has emerged as the big data platform of choice for data scientists both from the ease of use as well as the performance / optimization point of view. In a few lines of Scala code, Spark allows you to write iterative algorithms that scale out very well. For a data scientist who wants to explore large scale data sets, Spark is a great starting point (this is incredible progress in the Spark community given the project is just about 4 years old). However, Spark itself is moving fast and maturing with time, and Spark and Scala as well as distributed algorithms are typically not in the arsenal of many data scientists today.What this book does is teach you how to think about data science problems at scale, in the context of Spark. By well chosen examples covering both supervised and unsupervised learning, the authors take you step by step from a practical problem definition (say how to recommend music given user's history of music listened to) to what features are relevant, what machine learning algorithm to use and how to tune parameters to optimize the solution and how you can use Spark to do all of this in an interactive / iterative manner. As a bonus, they also point you to well engineered data sets that you can use to follow along the discussion and learn by trying out the examples yourself.By embracing the feature engineering steps and data cleaning/ error handling and tuning /feedback steps, the authors manage to show how real world data science works and how you can do full stack data science using Spark and gain immensely from the interactive nature of the Spark REPL.Overall, I highly recommend this book, and though it is the first book on Data Science using Spark, it sets a high standard for subsequent efforts.

fantastic book after you've learned the basics on your own. great detail, great examples, great explanation, great code. i read it by the pool while on vacation, which yes... makes me a nerd... but it gave me a solid next step after separate intros to Spark, MLlib, and Scala on my own.