8. Applying Decision Trees

  1. Phillip I. Good

Published Online: 23 FEB 2011

DOI: 10.1002/9780470937273.ch8

Analyzing the Large Number of Variables in Biomedical and Satellite Imagery

Analyzing the Large Number of Variables in Biomedical and Satellite Imagery

How to Cite

Good, P. I. (2011) Applying Decision Trees, in Analyzing the Large Number of Variables in Biomedical and Satellite Imagery, John Wiley & Sons, Inc., Hoboken, NJ, USA. doi: 10.1002/9780470937273.ch8

Author Information

  1. Huntington Beach CA, USA

Publication History

  1. Published Online: 23 FEB 2011
  2. Published Print: 21 MAR 2011

ISBN Information

Print ISBN: 9780470927144

Online ISBN: 9780470937273

SEARCH

Keywords:

  • biological data set;
  • decision tree;
  • ensemble methods;
  • maximally diversified multiple trees;
  • medical data sets;
  • receiver operating characteristic (ROC) analysis

Summary

The assignment of classifications to very large biological and medical data sets require three essential steps: standardization of the images so all have a common orientation and scale, reduction of variables, and classification. This chapter considers the methods that have been employed when classifying galactic images, sonographs, MRI, EEGs, MEG, mass spectral data, and microarrays. It describes a receiver operating characteristic (ROC) analysis based on plotting the true positive rate (TPR) on the y-axis and the false positive rate (FPR) on the x-axis. The efficiency of a decision tree can be increased by reducing the number of variables to be considered, increasing the size of the training set, correcting for relative costs, and correcting for the expected proportions in the target population.

Controlled Vocabulary Terms

ensemble forecasting; influence diagram; receiver operating characteristic curve