Book Image

Apache Spark 2: Data Processing and Real-Time Analytics

By : Romeo Kienzler, Md. Rezaul Karim, Sridhar Alla, Siamak Amirghodsi, Meenakshi Rajendran, Broderick Hall, Shuen Mei
Book Image

Apache Spark 2: Data Processing and Real-Time Analytics

By: Romeo Kienzler, Md. Rezaul Karim, Sridhar Alla, Siamak Amirghodsi, Meenakshi Rajendran, Broderick Hall, Shuen Mei

Overview of this book

Apache Spark is an in-memory, cluster-based data processing system that provides a wide range of functionalities such as big data processing, analytics, machine learning, and more. With this Learning Path, you can take your knowledge of Apache Spark to the next level by learning how to expand Spark's functionality and building your own data flow and machine learning programs on this platform. You will work with the different modules in Apache Spark, such as interactive querying with Spark SQL, using DataFrames and datasets, implementing streaming analytics with Spark Streaming, and applying machine learning and deep learning techniques on Spark using MLlib and various external tools. By the end of this elaborately designed Learning Path, you will have all the knowledge you need to master Apache Spark, and build your own big data processing and analytics pipeline quickly and without any hassle. This Learning Path includes content from the following Packt products: • Mastering Apache Spark 2.x by Romeo Kienzler • Scala and Spark for Big Data Analytics by Md. Rezaul Karim, Sridhar Alla • Apache Spark 2.x Machine Learning Cookbook by Siamak Amirghodsi, Meenakshi Rajendran, Broderick Hall, Shuen MeiCookbook
Table of Contents (23 chapters)
Title Page
Copyright
About Packt
Contributors
Preface
Index

CrossValidation and hyperparameter tuning


We will be looking at one example each of CrossValidation and hyperparameter tuning. Let's take a look at CrossValidation.

CrossValidation

As stated before, we've used the default parameters of the machine learning algorithm and we don't know if they are a good choice. In addition, instead of simply splitting your data into training and testing, or training, testing, and validation sets, CrossValidation might be a better choice because it makes sure that eventually all the data is seen by the machine learning algorithm.

Note

CrossValidation basically splits your complete available training data into a number of k folds. This parameter k can be specified. Then, the whole Pipeline is run once for every fold and one machine learning model is trained for each fold. Finally, the different machine learning models obtained are joined. This is done by a voting scheme for classifiers or by averaging for regression.

The following figure illustrates ten-fold CrossValidation...