#                                                                     
# R code to load in the vowel data set from the book ESLII and classify
# two classes using the routine
#
# two_class_LDA_with_optimal_cut_point
#
# and compare this routine to normal LDA (where the cut point is not specified optimally).
#
# See the section entitled:
#
# Linear Discriminant Analysis in ESLII 
#                                                                     
# Output:                                                             
#                                                                     
# res: list of data frames XT                                         
#                                                                     
# Written by:                                                         
# --                                                                  
# John L. Weatherwax                2009-04-21                        
#                                                                     
# email: wax@alum.mit.edu                                             
#                                                                     
# Please send comments and especially bug reports to the              
# above email address.                                                
#                                                                     
#-----                                                                

library(MASS) # this has functions for lda and qda 

source("load_vowel_data.R")
source("two_class_LDA_with_optimal_cut_point_pair_vowel.R")

out         = load_vowel_data( FALSE )
XTrain      = out[[1]]
labelsTrain = out[[2]] 
XTest       = out[[3]]
labelsTest  = out[[4]] 

# select a pair vowels to classify:
indexClassOne = 7
indexClassTwo = indexClassOne+1
two_class_LDA_with_optimal_cut_point_pair_vowel( XTrain,labelsTrain,XTest,labelsTest, indexClassOne,indexClassTwo )

num_of_experiments = 0 
num_times_optimal_cut_point_wins = 0
for( ii in min(labelsTrain):(max(labelsTrain)-1) ){
  for( jj in (ii+1):max(labelsTrain) ){
    new_method_better = two_class_LDA_with_optimal_cut_point_pair_vowel( XTrain,labelsTrain,XTest,labelsTest, ii, jj )
    num_of_experiments = num_of_experiments + 1 
    if( new_method_better ){
      num_times_optimal_cut_point_wins = num_times_optimal_cut_point_wins + 1 
    }
  }
}


print( sprintf("fraction of time optimal cut point is better than the orignal LDA= %10.6f", num_times_optimal_cut_point_wins / num_of_experiments ) );