The Evolutionary Pre-Processor (version 3.0)
Batch Run Report

 

Batch Details
Batch run name australian_2
Description Australian 2 on Nyquist
This report file final/australian_2/australian_2_report.html
data file /home/cssip/jsherrah/EPrep3/data/australian_2.dat
Time of completion Wed May 27 13:23:49 1998
Duration of Batch 2 hours, 4 minutes, 16 seconds
Random Seed 33564673 (from clock)
Average generations per run 11.40
Average failed feature creations per run 5548.00
Average fitness evalutions per run 55647.40
Test Set Improvement 22.54 %
 
 

Data Partition
 
Class Training Validation Test Total
0 192 95 96 383
1 153 77 77 307
Total 345 172 173 690

Summary of Results

  Original Classification Errors (%)
Classifier Training Validation Test
Parallelepiped(PPD) 35.36 37.79 39.31
Min. Distance to Means(MDTM) 34.78 38.37 38.73

  EPrep Best-of-run Classification Errors (%)
Run
Training
Validation
Test
McNemar confidence
# Features
# Inputs
# Nodes
Classifier
1 13.33 11.63 16.76 1.000 5 9 55 MDTM
2 13.91 12.79 17.34 1.000 1 5 14 MDTM
3 13.91 12.79 17.34 1.000 1 1 1 MDTM
4 13.91 12.79 17.34 1.000 1 1 1 MDTM
5 13.91 12.79 17.34 1.000 1 1 1 MDTM
6 13.91 12.79 17.34 1.000 1 2 12 MDTM
7 13.91 12.79 17.34 1.000 1 1 2 MDTM
8 14.20 12.79 17.34 1.000 1 3 9 MDTM
9 13.62 12.21 17.34 1.000 5 9 67 MDTM
10 12.46 11.63 16.18 1.000 3 7 43 MDTM
Ave. 13.71 (0.47 ) 12.50 (0.47 ) 17.17 (0.37 ) 1.00 (0.00 ) 2.00 (1.61 ) 3.90 (3.18 ) 20.50 (23.64 ) MDTM

  Confusion matrix for Best Ever Individual from run 1
Class
Predicted
Total
1
2
Ground Truth
1
78 18 96
2
11 66 77
Total
89 84 173

Average Operator Probabilities
Operator Average Probability
Delete-Feature Mutation 0.098
Add-Feature Mutation 0.100
Hoist Mutation 0.100
Truncate Mutation 0.100
Swap Mutation 0.100
One-Symbol Mutation 0.101
All-Nodes Mutation 0.099
One-Node Mutation 0.102
Grow Mutation 0.102
High-Level Crossover 0.100
 

Number of Run Terminations attributed to each Criterion
Termination Criterion Number of Terminations
TP Criterion 0
GL Criterion 10
Max. Generations 0
Client Abort 0
Zero Validation Error 0
Total 10
 
 

Related Data Files
Description Filename
australian_2_gen.m Matlab plot generation function
australian_2_bogf_ave.{eps,gif} Best-of-generation Fitness, averaged over 10 runs
australian_2_bogv_ave.{eps,gif} Best-of-generation Validation Set Error, averaged over 10 runs
australian_2_avef_ave.{eps,gif} Average fitness, averaged over 10 runs
australian_2_stdf_ave.{eps,gif} Standard deviation of fitness, averaged over 10 runs
australian_2_nftr_ave.{eps,gif} Average number of features per individual, averaged over 10 runs
australian_2_nnode_ave.{eps,gif} Average number of nodes per individual, averaged over 10 runs
australian_2_nint_ave.{eps,gif} Average number of introns per individual, averaged over 10 runs
australian_2_ntrl_ave.{eps,gif} Average number of RAT trials per individual, averaged over 10 runs
australian_2_optimp_ave.{eps,gif} Average improvement in fitness due to optimisation, averaged over 10 runs
australian_2_opprob_ave.{eps,gif} Average probability of each genetic operator, averaged over 10 runs
australian_2_run_x.dat Binary data file containing results of run x (read by Matlab functions)
australian_2_bor_x.prep 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-run individual for run x
australian_2_run_x.corr 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Feature correlation file for run x
australian_2_tst_bor_x.pred 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Test set predictions for best-of-run individual from run x
australian_2_bogf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Fitness for run x
australian_2_bogv_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Validation Set Error for run x
australian_2_avef_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average fitness for run x
australian_2_stdf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Standard deviation of fitness for run x
australian_2_nftr_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of features per individual for run x
australian_2_nnode_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of nodes per individual for run x
australian_2_nint_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of introns per individual for run x
australian_2_ntrl_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average number of RAT trials per individual for run x
australian_2_optimp_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average improvement in fitness due to optimisation for run x
australian_2_opprob_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average probability of each genetic operator for run x