The original paper compared the following 10 algorithms

SVMs:                            Linear, polynomial degree with radial width (.001, .005, .01, .05, .1, .5, 1, 2). 
ANN:                             Experiment with different values of hidden units(1,2,4,8,32,128) and momentum (0,0.2,0.5,0.9)
Logistic Regresion:        Ridge parameter is varied by factors of 10 from 10 power -8 to 10 power 3
Naive Bayes:
KNN:                             The value of k is varied from 1 to |trainset|
Random Forests:            Size of the feature set considered at each split is 1,2,4,6,8,12,16 or 20
Decision Trees:               BAYES, ID3, CART, CART0, C4, MML, and SMML
Bagged Trees:
Boosted Trees:               2,4,8,16,32,64,128,256,512,1024 and 2048 steps of boosting
Boosted Stumps:             2,4,8,16,32,64,128,256,512,1024,2048,4096,8192 steps


These algorithms were compared over the following 10 data sets

ADULT, COVER_TYPE, LETTER (UCI repositories)    
MEDIS, MG (Medical data sets)                                    
SLAC (Stanford linear accelerator)                                                                            
HS(Indian pine92) (UCI repositories)                     
CODING,BACTERIA,CALHOUS (UCI repositories)                                        

The nominal attributes in these data sets were converted to boolean for ANNs, KNN, SVM and Logistic regression


continue
Back to Project Home Page
Back to Varun's Home Page