Data Mining: Practical Machine Learning Tools and Techniques with Java Implementations (The Morgan Kaufmann Series in Data Management Systems) (Paperback)

by Ian H. Witten (Author), Eibe Frank (Author) "What is meant by ""structural patterns""?..." (more)
Key Phrases: informational loss function, basic covering algorithm, category utility formula, Naive Bayes, Peter Peggy, Pam Ian (more...)

Editorial Reviews

Amazon.com Review
Data mining techniques are used to power intelligent software, both on and off the Internet. Data Mining: Practical Machine Learning Tools explains the magic behind information extraction in a book that succeeds at bringing the latest in computer science research to any IS manager or developer. In addition, this book provides an opportunity for the authors to showcase their powerful reusable Java class library for building custom data mining software.

This text is remarkable with its comprehensive review of recent research on machine learning, all told in a very approachable style. (While there is plenty of math in some sections, the authors' explanations are always clear.) The book tours the nature of machine learning and how it can be used to find predictive patterns in data comprehensible to managers and developers alike. And they use sample data (for such topics as weather, contact lens prescriptions, and flowers) to illustrate key concepts.

After setting out to explain the types of machine learning models (like decision trees and classification rules), the book surveys algorithms used to implement them, plus strategies for improving performance and the reliability of results. Later the book turns to the authors' downloadable Weka (rhymes with "Mecca") Java class library, which lets you experiment with data mining hands-on and gets you started with this technology in custom applications. Final sections look at the bright prospects for data mining and machine learning on the Internet (for example, in Web search engines).

Precise but never pedantic, this admirably clear title delivers a real-world perspective on advantages of data mining and machine learning. Besides a programming how-to, it can be read profitably by any manager or developer who wants to see what leading-edge machine learning techniques can do for their software. --Richard Dragan

Topics covered: Data mining and machine learning basics, sample datasets and applications for data mining, machine learning vs. statistics, the ethics of data mining, generalization, concepts, attributes, missing values, decision tables and trees, classification rules, association rules, exceptions, numeric prediction, clustering, algorithms and implementations in Java, inferring rules, statistical modeling, covering algorithms, linear models, support vector machines, instance-based learning, credibility, cross-validation, probability, costs (lift charts and ROC curves), selecting attributes, data cleansing, combining multiple models (bagging, boosting, and stacking), Weka (reusable Java classes for machine learning), customizing Weka, visualizing machine learning, working with massive datasets, text mining, and e-mail and the Internet.

Review
"This is a milestone in the synthesis of data mining, data analysis, information theory and machine learning."
-Jim Gray, Microsoft Research, USA -- Review

See all Editorial Reviews

Product Details

• Paperback: 416 pages
• Publisher: Morgan Kaufmann; 1st edition (October 11, 1999)
• Language: English
• ISBN-10: 1558605525
• ISBN-13: 978-1558605527
• Product Dimensions: 9.1 x 7.4 x 0.8 inches
• Shipping Weight: 1.5 pounds
• Average Customer Review: 4.1 out of 5 stars See all reviews (17 customer reviews)
• Amazon.com Sales Rank: #754,032 in Books (See Bestsellers in Books)

  Popular in this category: (What's this?)

  #39 in

  Books > Computers & Internet > Databases > Java & Databases

Customer Reviews

Most Helpful Customer Reviews

35 of 36 people found the following review helpful:

5.0 out of 5 stars Excellent introduction to data mining algorithms, February 7, 2000

By	Dean (San Diego, CA United States) - See all my reviews

Witten and Frank have generated a book that is readable without eliminating all technical (yes, even mathematical!) descriptions of the key data mining algorithms. And they are up-to-date, including support vector machines and boosting. There are sufficient examples of the techniques to provide readers with a good feel for what each technique can accomplish. For example, how many books can provide a readable explanation of support vector machines?

There are some quibbles, such as not including any discussion of neural networks (noted in Ch. 1 with another reference)--I believe it deserves some attention because of its widespread use. Additionally, future editions should include a least a brief summary of data preprocessing, input selection, feature creation, etc. But these are quibbles.

The Java portion of the book is not of as much interest to me, but for those wishing to implement the algorithms, it provides a nice blueprint (from the code I looked at).

For what they have undertaken, they have performed admirably, and I would highly recommend this book.

26 of 27 people found the following review helpful:

5.0 out of 5 stars You HAVE to read this book!, January 28, 2000

By	Bostjan Brumen (Tampere University of Technology at Pori, Finland & University of Maribor, Slovenia) - See all my reviews

This book is THE best book I have read about data mining. And I have read most of them (see ISBNs: 0070057796, 0471253847, 0262560976, 0201403803, 0471179809, 013743980, 0137564120, 1558605290, 1558604030). It is fresh, clear, well balanced. If your native language is not English, then you should definetly read THIS book first.

The feature that is the most important for me is "just enough statistics". That is, you can understand the processes & descriptions even if you have not wasted your life and youth studying statistics; what is needed of it to understand is given shortly and very well. Many other books are too deep or too shallow (like Berry's, which is a good introduction, but nothing more than that).

If the rating was scaled 1-6 stars, I'd give this book a 10.

26 of 27 people found the following review helpful:

5.0 out of 5 stars Excellent data mining textbook, December 3, 1999

By	Stan Matwin (Ottawa, canada) - See all my reviews

Broad coverage, including hot new topics: SVM, boosting and bagging, modern evaluation methods (ROC and lift curves). Well grounded in practical data mining applications, talks about DM issues outside model building, which are rarely discussed: feature engineering, data cleaning, etc. Clear and well written: illustrative examples help the presentation a lot. Describes in detail decision trees and rule learners, instance-based learning, and numerical prediction. Accompanied by the WEKA system, implementing in Java many of the methods discussed in the book, and available for download for free. An excellent hands-on textbook for an applied Machine Learening/DM class, or recommended reading for ayone who wants to understand DM. Good next step for those that have whetted their appetite with Berry and Linof's book.