Preface

There are many notable books on machine learning, from pedagogical tracts on the theory of learning from data; to standard references on specializations in the field, such as clustering and outlier detection or probabilistic graph modeling; to cookbooks that offer practical advice on the use of tools and libraries in a particular language. The books that tend to be broad in coverage are often short on theoretical detail, while those with a focus on one topic or tool may not, for example, have much to say about the difference in approach in a streaming as opposed to a batch environment. Besides, for the non-novices with a preference for tools in Java who wish to reach for a single volume that will extend their knowledge—simultaneously, on the essential aspects—there are precious few options.

Finding in one place

• The pros and cons of different techniques given any data availability scenario—when data is labeled or unlabeled, streaming or batch, local, or distributed, structured or unstructured
• A ready reference for the most important mathematical results related to those very techniques for a better appreciation of the underlying theory
• An introduction to the most mature Java-based frameworks, libraries, and visualization tools with descriptions and illustrations on how to put these techniques into practice is not possible today, as far as we know

The core idea of this book, therefore, is to address this gap while maintaining a balance between treatment of theory and practice with the aid of probability, statistics, basic linear algebra, and rudimentary calculus in the service of one, and emphasizing methodology, case studies, tools and code in support of the other.

According to the KDnuggets 2016 software poll, Java, at 16.8%, has the second highest share in popularity among languages used in machine learning, after Python. What's more is that this marks a 19% increase from the year before! Clearly, Java remains an important and effective vehicle to build and deploy systems involving machine learning, despite claims of its decline in some quarters. With this book, we aim to reach professionals and motivated enthusiasts with some experience in Java and a beginner's knowledge of machine learning. Our goal is to make Mastering Java Machine Learning the next step on their path to becoming advanced practitioners in data science. To guide them on this path, the book covers a veritable arsenal of techniques in machine learning—some which they may already be familiar with, others perhaps not as much, or only superficially—including methods of data analysis, learning algorithms, evaluation of model performance, and more in supervised and semi-supervised learning, clustering and anomaly detection, and semi-supervised and active learning. It also presents special topics such as probabilistic graph modeling, text mining, and deep learning. Not forgetting the increasingly important topics in enterprise-scale systems today, the book also covers the unique challenges of learning from evolving data streams and the tools and techniques applicable to real-time systems, as well as the imperatives of the world of Big Data:

• How does machine learning work in large-scale distributed environments?
• What are the trade-offs?
• How must algorithms be adapted?
• How can these systems interoperate with other technologies in the dominant Hadoop ecosystem?

This book explains how to apply machine learning to real-world data and real-world domains with the right methodology, processes, applications, and analysis. Accompanying each chapter are case studies and examples of how to apply the newly learned techniques using some of the best available open source tools written in Java. This book covers more than 15 open source Java tools supporting a wide range of techniques between them, with code and practical usage. The code, data, and configurations are available for readers to download and experiment with. We present more than ten real-world case studies in Machine Learning that illustrate the data scientist's process. Each case study details the steps undertaken in the experiments: data ingestion, data analysis, data cleansing, feature reduction/selection, mapping to machine learning, model training, model selection, model evaluation, and analysis of results. This gives the reader a practical guide to using the tools and methods presented in each chapter for solving the business problem at hand.