Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Feature selection for data and pattern recognition
Stanczyk U., Jain L., Springer Publishing Company, Incorporated, New York, NY, 2014. 355 pp. Type: Book (978-3-662456-19-4)
Date Reviewed: Jun 16 2015

Feature selection is one of the most important preprocessing steps, with the performance of any system designed to solve pattern recognition, or data mining tasks in general, being strongly dependent on the quality of the feature set in terms of which processed objects are represented. Stated in general terms, a feature selection method is a means of extracting relevant information from data with respect to the particular task to be solved, described either by a given model or, as is the case with real-world problems, by a finite collection of examples.

The book is a collection of 14 research texts structured into four parts written by several representative scientists in the field, supplying the reader with a comprehensive and sound presentation of the most recent and advanced developments, as well as the main trends in feature selection methodologies for pattern recognition purposes. The first two chapters offer a brief overview of the feature selection problem and the most frequently used methods and applications. Due to the high dimensionality of object representations in terms of all possible measurable or observable descriptors, any classification problem can become too expensive from a computational point of view, or even intractable; therefore, the evaluation of the efficiency corresponding to different candidate subsets of features, as well as the use of dimensionality reduction methods, is a crucial step in implementing a feature selection process.

The third chapter of the book illustrates a two-step processing framework where filter, wrapper, and embedded approaches are combined in order to develop a suitable feature selection model. A new method, referred to as feature ranking (FR), based on the computation of the relative weights of features and a series of feature ranking methods, is proposed in chapter 4. Chapter 5 investigates the problem of weighting features by procedures of sequential backward and forward selection in the framework of neural network-based approaches.

The next two chapters are focused on rough set approaches for attribute reduction purposes. Two probabilistic approaches to rough sets for different types of data dependency detection and their representations, namely the variable precision rough set model and the Bayesian rough set model, are presented in the sixth chapter, followed by a detailed theoretical development of a rough set approach for attribute reduction purposes in the next chapter.

The next three chapters, comprising the third part of the book, are devoted to rule extraction and evaluation. The results of a series of tests aiming to develop a comparative analysis between the performances of the learning from examples module, version 2 (LEM2) rule induction algorithm and the LEM1 algorithm are reported in chapter 8. The results lead to the conclusion that the LEM2 algorithm with the space search of all attribute-value pairs performs better than LEM1 based on feature selection. In the next two chapters, the benefits of meta-actions in evaluating action rules are investigated in terms of likelihood and execution confidence, and the application of several statistical methods to optimize the subtree-based associative classification in the case of tree-structured data, respectively.

The final four chapters, comprising the fourth part of the book, present different aspects related to data- and domain-oriented methodologies. Chapter 11 explores the problem of data classification in cases where the data containing instances similar to a large number of other instances, conventionally referred to as hubs, are taken into account, with an integrated framework from the point of view of notations and terminology. The aim of the next chapter is to develop a comparative analysis of different visual descriptors of image data, with the results of several tests discussed in detail. The final two chapters deal with the use of a genetic algorithm and the relief algorithm to select features when discriminating among human faces and facial expressions, and methods of group-wise feature selection in addition to individual features when features are grouped, respectively.

The content of the book is outstanding from the point of view of the novelty of the exposed methods, the clarity of the discourse, and the variety of the illustrative examples. Each chapter contains a list of bibliographic references containing the most representative titles in the field. The book is aimed at researchers and practitioners in the domains of machine learning, computer science, data mining, statistical pattern recognition, and bioinformatics.

Reviewer:  L. State Review #: CR143527 (1509-0773)
Bookmark and Share
  Reviewer Selected
 
 
Feature Evaluation And Selection (I.5.2 ... )
 
 
Applications (I.5.4 )
 
 
Implementation (I.5.5 )
 
Would you recommend this review?
yes
no
Other reviews under "Feature Evaluation And Selection": Date
Labeled point pattern matching by Delaunay triangulation and maximal cliques
Ogawa H. Pattern Recognition 19(1): 35-40, 1986. Type: Article
Feb 1 1988
Features selection and ‘possibility theory’
Di Gesù V., Maccarone M. Pattern Recognition 19(1): 63-72, 1986. Type: Article
Dec 1 1987
An analytic-to-holistic approach for face recognition based on a single frontal view
Lam K., Yan H. IEEE Transactions on Pattern Analysis and Machine Intelligence 20(7): 673-686, 1998. Type: Article
Oct 1 1998
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy