MULTIVARIATE TECHNIQUE FOR CLASSIFICATION RULE SEARCHING - Exemplieied by CT Data of Patient

Jyhjeng Deng

2008

Abstract

In the process of searching classification rules for multivariate categorical data, it is crucial to find a quick start to locate the combination of levels of input and response variables which can contribute to the most correct classification rate for the response variable. Fisher’s linear discriminant function is proposed to select some important input-variable candidates; then, correspondence analysis is used to ascertain that the level of candidates is closely related to the appropriate level of response variable. The closest linkage between input variable and response variables is chosen as the rule for each input-variable candidate. The algorithm is applied to the hospital data of patients whose CT scan diagnosis awaits a decision. The result shows that my algorithm is not only quicker than an exhaustive search but the result is also identical to the optimum solution by exhaustive search in terms of the correct classification rate. The correct classification rate is about 80%. Finally, two parallel coordinate plots of the 20% mistakenly classified data and the corresponding correctly classified data are compared, showing their mutual confounding and explaining why the correct classification rate cannot be further improved.

References

  1. Yang, M., Kriegman, D and Ahuja, N., 2001, Face Detection Using Multimodal Density Models, Computer Vision and Image Understanding 84, 264- 284.
  2. Lam, K. and Moy, J., 2003, A piecewise linear programming approach to the two-group discriminant Operational Research 145, 471- 481.
  3. Allombert, S., Gaston, A. and Martin, J., 2005, A natural experiment on the impact of overabundant deer on songbird populations, Biological Conservation 126, 1- 13.
  4. Hardle, W. and Simar L., 2003. Applied Multivariate Statistical Analysis, Springer. Berlin.
  5. Johnson R. and Wichern D., 2002, Applied Multivariate Statistical Analysis, Prentice Hall. 5th, NJ, USA.
  6. Hardle, W., Klinke, S. and Muller M., 2000. XploRe Learning Guide, Springer. Berlin.
Download


Paper Citation


in Harvard Style

Deng J. (2008). MULTIVARIATE TECHNIQUE FOR CLASSIFICATION RULE SEARCHING - Exemplieied by CT Data of Patient . In Proceedings of the Tenth International Conference on Enterprise Information Systems - Volume 4: CIAS, (ICEIS 2008) ISBN 978-989-8111-39-5, pages 278-286. DOI: 10.5220/0001723702780286


in Bibtex Style

@conference{cias08,
author={Jyhjeng Deng},
title={MULTIVARIATE TECHNIQUE FOR CLASSIFICATION RULE SEARCHING - Exemplieied by CT Data of Patient},
booktitle={Proceedings of the Tenth International Conference on Enterprise Information Systems - Volume 4: CIAS, (ICEIS 2008)},
year={2008},
pages={278-286},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0001723702780286},
isbn={978-989-8111-39-5},
}


in EndNote Style

TY - CONF
JO - Proceedings of the Tenth International Conference on Enterprise Information Systems - Volume 4: CIAS, (ICEIS 2008)
TI - MULTIVARIATE TECHNIQUE FOR CLASSIFICATION RULE SEARCHING - Exemplieied by CT Data of Patient
SN - 978-989-8111-39-5
AU - Deng J.
PY - 2008
SP - 278
EP - 286
DO - 10.5220/0001723702780286