Results for classifiers learnt from 20 initial configurations.

========
TABLE II
========

model type           accuracy    cross-validation     sensitivity    specificity
SVM-linear            44.00%          52.07%             51.01%         37.06%
SVM-polynomial        58.09%          53.07%             63.04%         53.14%
SVM-RBF               94.51%          94.79%             99.09%         89.94%



=========
TABLE III
=========

number of mutants        effective mutants         accuracy           cross-validation      
300                             40                  85.01%               83.56%   
400                             61                  89.18%               90.01%
500                             79                  93.13%               93.59%
600                             98                  94.13%               94.77%
700                             108                 94.51%               94.79%


========
TABLE IV
========
#time interval       accuracy     cross-validation
100                   94.01%          93.68%
150                   94.73%          92.17%
200                   93.47%          94.67%
250                   94.51%          94.79%
300                   94.54%          95.51%
=======
TABLE V
=======

attack #             detected         accuracy
1                       yes            93.47%
2                       yes            91.11%
3                       yes            75.76%   
4                       yes            99.19%       
5                       yes            96.28%
6                       yes            94.42%
7                       yes            65.19%
8                       yes            96.09%
9                       yes            91.91%
10                      yes            93.17%
11                      yes            95.04%
12                      yes            96.18%
13                      yes            95.74%
14                      yes            94.57%
15                      yes            96.66%
           
========
TABLE VI
========
attack stage   # effective mutants        # detected     accuracy (detected)    accuracy (all)
PLC1                    8                      5               99.31%               73.14%    
PLC3                   20                     17               99.81%               90.91%
PLC4                    5                      4               98.15%               92.07%
PLC5                    2                      2               99.54%               98.88%                     
PLC6                    5                      5               99.00%               99.07%
summary                40                     33               99.85%               90.17%