Advertisement

Hybrid GA-SVM-Oracle Paradigm

  • Walker H. LandJr.
  • J. David Schaffer
Chapter
  • 325 Downloads

Abstract

In this chapter, we show how the methods described in previous chapters can be effectively combined. The application is the supervised learning of an accurate classifier. An example is characterized by a vector of features, and one presumes more features than needed, so some form of feature subset selection is needed. We describe what is often called a “wrapper approach” in which the feature subset selection algorithm (CHC in this case) repeatedly calls a leaning classifier system (SVM in this case) to compute an evaluation of each proffered feature set. We then describe an approach to data normalization that allows new cases that need to be classified in future, to have feature values outside the range observed in the training set. The two measures of desired performance, classification accuracy and subset size, are shown both evolving in the desired directions. We introduce an extension of the conventional area under the ROC curve metric (Sect.  3.7) that has desirable properties for tasks of this type. Then we apply this hybrid approach to our task of discovering a speech-based diagnostic test for dementia including aspects of the cross-validation experimental design and schemes for picking the winning feature sets. Finally, we show how the multiple discovered feature sets may be effectively combined using the GRNN oracle. This revealed some cases that we call trouble-makers, discussed in Chap.  3. These provide examples for the new method presented in Chap.  9 that allows a classifier system to know what it doesn’t know.

Keywords

Cross-validation Data normalization Extension of area under ROC Picking the winners CHC algorithm Feature subset selection GRNN Oracle 

Abbreviations

AD

Alzheimer’s disease

AUC

Area under the ROC curve

CHC

Eshelman’s GA

EEG

Electroencephalogram

GA

Genetic algorithm

GRNN

Generalized regression neural network

LIWC

Lexical inquiry word count

MMSE

Mini Mental State Exam

MMX_SSS

Match and mix crossover operator for subset selection tasks

NL

Cognitively normal

ROC

Receiver operator characteristic

SVM

Support vector machine

VAD

Voice activity detector

References

  1. Pennebaker JW, Boyd RL, Jordan K, Blackburn K (2015) The development and psychometric properties of LIWC2015. University of Texas at Austin, AustinGoogle Scholar
  2. Roark B, Mitchell M, Hosom J, Hollingshead K, Kaye J (2011) Spoken language derived measures for detecting mild cognitive impairment. IEEE Trans Audio Speech Lang Process 19(7):2081–2090CrossRefGoogle Scholar
  3. Bar Siman Tov O, Schaffer JD, McLeod K (2015) Developing an evolutionary algorithm to search for an optimal multi-mother wavelet packets combination. J Biomed Sci Eng 8(7):458–470CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  • Walker H. LandJr.
    • 1
  • J. David Schaffer
    • 2
  1. 1.Binghamton UniversityBowieUSA
  2. 2.Binghamton UniversityBinghamtonUSA

Personalised recommendations