DM: FYI: Empirical comparison of classification trees with missing valu

[Tjen-Sien Lim]

I've updated my empirical comparison of prediction accuracy of
classification trees and rules induction algorithms on data sets with
missing covariate values. The following results are averages from 34
data sets. More details are available at

     http://www.stat.wisc.edu/~limt/treeprogs.html

Thanks for your attention.


     Methods                 Means Rank       Means Error Rate (%)
==================================================================

 1   Boosted C5.0 Rules          6.7            21.20
 2   Boosted C5.0 Tree           8.1            21.80  
 3   Bagged CARTŪ (r)              8.7            21.96 
 4   M5class                    10.3            22.97
 5   ARCed CARTŪ (r)              10.8            22.38 
 6   QUEST (linear, 0-SE)       11.1            22.91
 7   CARTŪ (r)                    11.2            23.41
 8   Boosted-RIPPER             11.5            22.67 
 9   IND CARTŪ  (0-SE)            12.6            23.76 
10   IND MML options            13.3            24.74
11   IND MML                    13.8            24.47 
12   IND CARTŪ  (1-SE)            13.8            24.15
13   QUEST (univariate, 0-SE)   14.0            24.54
14   C5.0 Rules                 15.1            24.53
15   IND Bayes options          15.4            24.59 
16   Ripley's Tree function     15.8            25.61
17   C5.0 Tree                  15.8            24.72 
18   C4.5 Tree                  16.2            24.77 
19   RPART                      17.2            26.10
20   C4.5 Rules                 17.3            25.10
21   RIPPER                     17.4            25.85
22   IND Bayes                  17.4            24.89  
23   CN2                        18.2            26.14  
24   KnowledgeSEEKER            19.0            26.17 
25   S-Plus Tree                19.1            27.07 
26   ITI                        19.2            26.02
27   FIRM                       21.4            29.10
28   DMTI                       21.5            26.55
29   T1                         23.0            33.74


-- 
Tjen-Sien Lim                (608) 262-8181                        
Dept. of Statistics          limt@stat.wisc.edu
Univ. of Wisconsin-Madison   http://www.stat.wisc.edu/~limt
1210 West Dayton Street       
Madison, WI 53706