lung-cancer.names

1. Title: Lung Cancer Data

2. Source Information:
	- Data was published in : 
	  Hong, Z.Q. and Yang, J.Y. "Optimal Discriminant Plane for a Small
	  Number of Samples and Design Method of Classifier on the Plane",
	  Pattern Recognition, Vol. 24, No. 4, pp. 317-324, 1991.
	- Donor: Stefan Aeberhard, stefan@coral.cs.jcu.edu.au
	- Date : May, 1992

3. Past Usage:
	- Hong, Z.Q. and Yang, J.Y. "Optimal Discriminant Plane for a Small
          Number of Samples and Design Method of Classifier on the Plane",
          Pattern Recognition, Vol. 24, No. 4, pp. 317-324, 1991.
	- Aeberhard, S., Coomans, D, De Vel, O. "Comparisons of 
	  Classification Methods in High Dimensional Settings", 
	  submitted to Technometrics.
	- Aeberhard, S., Coomans, D, De Vel, O. "The Dangers of 
	  Bias in High Dimensional Settings", submitted to
	  pattern Recognition.

4. Relevant Information:
	- This data was used by Hong and Young to illustrate the 
	  power of the optimal discriminant plane even in ill-posed
	  settings. Applying the KNN method in the resulting plane	
	  gave 77% accuracy. However, these results are strongly
	  biased (See Aeberhard's second ref. above, or email to
	  stefan@coral.cs.jcu.edu.au). Results obtained by
	  Aeberhard et al. are : 
	  RDA : 62.5%, KNN 53.1%, Opt. Disc. Plane 59.4%

	  The data described 3 types of pathological lung cancers.
	  The Authors give no information on the individual
	  variables nor on where the data was originally used.

       -  In the original data 4 values for the fifth attribute were -1.
          These values have been changed to ? (unknown). (*)
       -  In the original data 1 value for the 39 attribute was 4.  This
          value has been changed to ? (unknown). (*)
    
	  
5. Number of Instances: 32

6. Number of Attributes: 57 (1 class attribute, 56 predictive)

7. Attribute Information:

	attribute 1 is the class label.
	
	- All predictive attributes are nominal, taking on integer 
	  values 0-3

8. Missing Attribute Values: Attributes 5 and 39 (*)

9. Class Distribution:
	- 3 classes, 
		1.)	9 observations
		2.)	13     "
		3.)	10     "