Instance Selection + C4.5 on Pima Indians Diabetes
# 17/11/2009 23:11
Test description:
Crosvalidation test 10x10 Normalization [-1,1] Vector Selection Classification=C45
C45 Configuration:
BinarySplits=False ConfidenceFactor=0.25 MinNumObj=2 NumFolds=3 ReducedErrorPruning=False Unpruned=False
Pima Indians Diabetes
| Accuracy | Compression | Rules | ||||
|---|---|---|---|---|---|---|
| mean | std. | mean | std. | mean | std. | |
| C4.5 | 73.23 | 4.11 | 100.00 | 0.00 | 7.71 | 1.71 |
| EkPs60p1 | 65.11  | 5.21 | 0.29 | 0.00 | 1.00 | 0.00 |
| EkPs60p3 | 73.72 | 4.12 | 0.87 | 0.00 | 2.00 | 0.00 |
| EkPs60p5 | 73.78 | 4.39 | 1.45 | 0.00 | 2.07 | 0.26 |
| EkPs60p10 | 73.46 | 4.25 | 2.89 | 0.00 | 2.75 | 0.64 |
| EkPs10p1 | 65.11 | 5.21 | 0.29 | 0.00 | 1.00 | 0.00 |
| CNN | 71.80 | 5.55 | 50.66 | 1.24 | 3.89 | 1.38 |
| DROP3 | 71.09 | 5.72 | 12.29 | 1.50 | 5.67 | 1.91 |
| DROP5 | 71.59 | 5.13 | 17.48 | 1.80 | 5.19 | 2.08 |
| ENRBF | 70.74 | 6.23 | 11.07 | 0.87 | 5.15 | 1.34 |
| ENN | 74.90 | 4.41 | 85.70 | 0.73 | 9.47 | 1.98 |
| ICF | 69.62 | 7.14 | 23.06 | 1.20 | 3.73 | 1.27 |
| GE | 73.53 | 4.07 | 94.62 | 0.58 | 7.40 | 1.61 |
| RNGE | 72.31 | 4.82 | 60.03 | 1.02 | 4.61 | 1.42 |
| RMHC | 65.11 | 5.21 | 0.29 | 0.00 | 1.00 | 0.00 |
| MC | 65.11 | 5.21 | 0.29 | 0.00 | 1.00 | 0.00 |
statistically significant improvement, significant degradation
EkPs60p1 Configuration:
Classifier=C45 PostClassifier=C45 CostTolerance=1E-16 MaxNumCostCalls=300 MaxSimplexIters=200 NumFolds=-1 NumProtoPerClass=1 NumSimplexPoints=60
EkPs60p3 Configuration:
Classifier=C45 PostClassifier=C45 CostTolerance=1E-16 MaxNumCostCalls=300 MaxSimplexIters=200 NumFolds=-1 NumProtoPerClass=3 NumSimplexPoints=60
EkPs60p5 Configuration:
Classifier=C45 PostClassifier=C45 CostTolerance=1E-16 MaxNumCostCalls=300 MaxSimplexIters=200 NumFolds=-1 NumProtoPerClass=5 NumSimplexPoints=60
EkPs60p10 Configuration:
Classifier=C45 PostClassifier=C45 CostTolerance=1E-16 MaxNumCostCalls=300 MaxSimplexIters=200 NumFolds=-1 NumProtoPerClass=10 NumSimplexPoints=60
EkPs10p1 Configuration:
Classifier=C45 PostClassifier=C45 CostTolerance=1E-16 MaxNumCostCalls=300 MaxSimplexIters=200 NumFolds=-1 NumProtoPerClass=1 NumSimplexPoints=60
CNN Configuration:
Classifier=kNN k=3, Metric=Intemi.Metric.SquareEuclidean
DROP3 Configuration:
Classifier=kNN k=3, Metric=Intemi.Metric.SquareEuclidean Order=EnemyDistance ENN=True IncludePruned=True
DROP5 Configuration:
Classifier=kNN k=3, Metric=Intemi.Metric.SquareEuclidean Order=EnemyDistanceAndReverse ENN=False IncludePruned=True
ENRBF Configuration:
Sigma=0.5 Strenght=0.05 RemoveNoise=True Tolerance=0.95 Metric=Intemi.Metric.SquareEuclidean
ENN Configuration:
Classifier=kNN k=3, Metric=Intemi.Metric.SquareEuclidean Repeated=False
ICF Configuration:
Metric=Intemi.Metric.SquareEuclidean Filter=(ENN using kNN k=3)
GE Configuration:
Metric=Intemi.Metric.SquareEuclidean MeighboursType=Gabriel
RNGE Configuration:
Metric=Intemi.Metric.SquareEuclidean MeighboursType=Relative
RMHC Configuration:
Classifier=kNN k=1, Metric=Intemi.Metric.SquareEuclidean Search=randomMutation Mutations=1000
MC Configuration:
Classifier=kNN k=1, Metric=Intemi.Metric.SquareEuclidean Search=monteCarlo Mutations=1000