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

 

Batch Details
Batch run name australian_3
Description Australian 3 on Nyquist
This report file final/australian_3/australian_3_report.html
data file /home/cssip/jsherrah/EPrep3/data/australian_3.dat
Time of completion Thu May 28 11:28:54 1998
Duration of Batch 1 hours, 43 minutes, 28 seconds
Random Seed 18446744072241555456 (from clock)
Average generations per run 15.20
Average failed feature creations per run 5464.70
Average fitness evalutions per run 63826.40
Test Set Improvement 20.81 %
 
 

Data Partition
 
Class Training Validation Test Total
0 191 97 95 383
1 154 75 78 307
Total 345 172 173 690

Summary of Results

  Original Classification Errors (%)
Classifier Training Validation Test
Parallelepiped(PPD) 37.97 41.28 36.99
Min. Distance to Means(MDTM) 37.39 34.88 31.79

  EPrep Best-of-run Classification Errors (%)
Run
Training
Validation
Test
McNemar confidence
# Features
# Inputs
# Nodes
Classifier
1 16.81 12.79 11.56 1.000 1 1 1 MDTM
2 17.68 13.37 11.56 1.000 2 11 97 MDTM
3 16.23 12.21 11.56 1.000 3 9 36 MDTM
4 16.52 12.21 11.56 1.000 2 9 43 MDTM
5 16.81 12.79 11.56 1.000 1 1 7 MDTM
6 17.10 11.63 10.98 1.000 1 2 5 MDTM
7 16.52 12.21 10.98 1.000 2 3 7 MDTM
8 16.81 12.79 11.56 1.000 2 3 16 MDTM
9 15.94 11.63 11.56 1.000 4 11 76 MDTM
10 14.20 12.21 12.14 1.000 4 10 63 MDTM
Ave. 16.46 (0.88 ) 12.38 (0.52 ) 11.50 (0.31 ) 1.00 (0.00 ) 2.20 (1.08 ) 6.00 (4.10 ) 35.10 (32.19 ) MDTM

  Confusion matrix for Best Ever Individual from run 6
Class
Predicted
Total
1
2
Ground Truth
1
83 12 95
2
7 71 78
Total
90 83 173

Average Operator Probabilities
Operator Average Probability
Delete-Feature Mutation 0.100
Add-Feature Mutation 0.096
Hoist Mutation 0.099
Truncate Mutation 0.100
Swap Mutation 0.101
One-Symbol Mutation 0.100
All-Nodes Mutation 0.102
One-Node Mutation 0.098
Grow Mutation 0.101
High-Level Crossover 0.104
 

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_3_gen.m Matlab plot generation function
australian_3_bogf_ave.{eps,gif} Best-of-generation Fitness, averaged over 10 runs
australian_3_bogv_ave.{eps,gif} Best-of-generation Validation Set Error, averaged over 10 runs
australian_3_avef_ave.{eps,gif} Average fitness, averaged over 10 runs
australian_3_stdf_ave.{eps,gif} Standard deviation of fitness, averaged over 10 runs
australian_3_nftr_ave.{eps,gif} Average number of features per individual, averaged over 10 runs
australian_3_nnode_ave.{eps,gif} Average number of nodes per individual, averaged over 10 runs
australian_3_nint_ave.{eps,gif} Average number of introns per individual, averaged over 10 runs
australian_3_ntrl_ave.{eps,gif} Average number of RAT trials per individual, averaged over 10 runs
australian_3_optimp_ave.{eps,gif} Average improvement in fitness due to optimisation, averaged over 10 runs
australian_3_opprob_ave.{eps,gif} Average probability of each genetic operator, averaged over 10 runs
australian_3_run_x.dat Binary data file containing results of run x (read by Matlab functions)
australian_3_bor_x.prep 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-run individual for run x
australian_3_run_x.corr 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Feature correlation file for run x
australian_3_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_3_bogf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Best-of-generation Fitness for run x
australian_3_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_3_avef_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average fitness for run x
australian_3_stdf_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Standard deviation of fitness for run x
australian_3_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_3_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_3_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_3_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_3_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_3_opprob_x.{eps,gif} 
x = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Average probability of each genetic operator for run x