Book Image

Scala for Data Science

By : Pascal Bugnion
Book Image

Scala for Data Science

By: Pascal Bugnion

Overview of this book

Scala is a multi-paradigm programming language (it supports both object-oriented and functional programming) and scripting language used to build applications for the JVM. Languages such as R, Python, Java, and so on are mostly used for data science. It is particularly good at analyzing large sets of data without any significant impact on performance and thus Scala is being adopted by many developers and data scientists. Data scientists might be aware that building applications that are truly scalable is hard. Scala, with its powerful functional libraries for interacting with databases and building scalable frameworks will give you the tools to construct robust data pipelines. This book will introduce you to the libraries for ingesting, storing, manipulating, processing, and visualizing data in Scala. Packed with real-world examples and interesting data sets, this book will teach you to ingest data from flat files and web APIs and store it in a SQL or NoSQL database. It will show you how to design scalable architectures to process and modelling your data, starting from simple concurrency constructs such as parallel collections and futures, through to actor systems and Apache Spark. As well as Scala’s emphasis on functional structures and immutability, you will learn how to use the right parallel construct for the job at hand, minimizing development time without compromising scalability. Finally, you will learn how to build beautiful interactive visualizations using web frameworks. This book gives tutorials on some of the most common Scala libraries for data science, allowing you to quickly get up to speed with building data science and data engineering solutions.
Table of Contents (22 chapters)
Scala for Data Science
Credits
About the Author
About the Reviewers
www.PacktPub.com
Preface
Index

Introducing MLlib – Spam classification


Let's introduce MLlib with a concrete example. We will look at spam classification using the Ling-Spam dataset that we used in the Chapter 10, Distributed Batch Processing with Spark. We will create a spam filter that uses logistic regression to estimate the probability that a given message is spam.

We will run through examples using the Spark shell, but you will find an analogous program in LogisticRegressionDemo.scala among the examples for this chapter. If you have not installed Spark, refer to Chapter 10, Distributed Batch Processing with Spark, for installation instructions.

Let's start by loading the e-mails in the Ling-Spam dataset. If you have not done this for Chapter 10, Distributed Batch Processing with Spark, download the data from data.scala4datascience.com/ling-spam.tar.gz or data.scala4datascience.com/ling-spam.zip, depending on whether you want a tar.gz file or a zip file, and unpack the archive. This will create a spam directory and...