Instant Weka How-to

by Boštjan Kaluža

Java code snippets for most commonly encountered ML tasks using Weka. It covers most of the ML tasks (classification, regression, association rules, clustering). Some case studies are also appended.

80 pages, Paperback

First published January 1, 2013

Reviews

[Lucian · Rating 4 out of 5]

Weka is one of the most popular software used for machine learning experiments, in Java world; the software is introduced by “Data Mining: Practical Machine Learning Tools and Techniques”. There are two approaches for using Weka: one is through the GUI and another one by programmatically calling the API. The book “Instant Weka how-to” targets programmers willing to integrate Weka inside their programs. It is written by a researcher with a proven experience in this area. In only 80 pages, the reader is shown some practical examples on how to embed Weka ML functionalities in Java applications.

What I like here is the brevity of exposition: for each example, the aims are clearly delineated; the main piece of code - together with imported packages - is shown as a whole block at first and then it is split into functional pieces.

The book starts with showing how to add reference to Weka jar into a Eclipse Java project (here the alternative of importing Weka library through Maven is unfortunately missing , but manually adding jar is a quicker approach for first examples). The programmer is then shown how to load an arff file, how to apply some preprocessing steps (in Weka parlance: filters), training a classifier, adding custom classifier, which is a plus), how to test and evaluate model through k-fold cross validation, how to produce confusion matrix and graphical representation of ROC curve, regression models, association rules, clustering and cluster evaluation, and (de)serialization of the models, among others.

Finally three practical DM problems are solved: classification (predicting buyer/non buyer), stock value forecasting and building a recommendation system.

Some more complex examples would make this books more appealing, e.g. applying a chain of filters (multifilter functionality from Weka),using cross validation for parameter selection, adding more details on how to create the plugin package.

The books fulfills the target stated by its title: a Java programmer can quickly embed Weka models inside his/her own code.

In a nut shell, the code snippets are clear, with no unnecessary burden, and the material presented is well delivered. In my opinion, the books is worth the price.

[Vasanth · Rating 4 out of 5]

Weka is undoubtedly the best open source Machine Learning library written in Java. Most people are comfortable using Weka as an easy to use GUI tool for their data analysis tasks. However, the Weka libraries can be also used from Java code by importing appropriate JAR files. There are very limited material available across the web in this regard. This book "Instant Weka How-To" delves deep into explaining different possibilities of using Weka libraries through Java code by providing examples. The examples covered in this book encompass the areas of classification, clustering, association rule mining, recommendation systems etc. This is a very good beginners guide for anybody with Java knowledge to get started with data analysis tasks without any hassles. I strongly recommend this book for wannabe data scientists, who likes to do hands-on experiments with data .