Detailed Results¶
Complete Result Tables¶
Iris Dataset (20-fold CV)¶
| Fold | GA Acc | CART Acc | GA Nodes | CART Nodes |
|---|---|---|---|---|
| 1 | 0.933 | 0.900 | 7 | 16 |
| 2 | 0.967 | 0.933 | 7 | 16 |
| ... | ... | ... | ... | ... |
| 20 | 0.933 | 0.967 | 9 | 18 |
| Mean | 0.9455 | 0.9241 | 7.4 | 16.4 |
| Std | 0.0807 | 0.1043 | 2.1 | 4.2 |
Statistical Test:
- t-statistic: 1.371
- p-value: 0.186 (not significant)
- Cohen's d: 0.230 (small effect)
Wine Dataset (20-fold CV)¶
| Metric | GA | CART |
|---|---|---|
| Accuracy | 88.19 ± 10.39% | 87.22 ± 10.70% |
| F1-Score | 87.89 ± 10.63% | 86.87 ± 10.99% |
| Nodes | 10.7 ± 3.1 | 20.7 ± 5.8 |
| Depth | 3.0 ± 0.8 | 4.4 ± 1.2 |
Statistical Test:
- p-value: 0.683 (not significant)
- Size reduction: 48%
Breast Cancer (20-fold CV)¶
| Metric | GA | CART |
|---|---|---|
| Accuracy | 91.05 ± 5.60% | 91.57 ± 3.92% |
| Nodes | 6.5 ± 2.1 | 35.5 ± 4.2 |
| Depth | 2.3 ± 0.9 | 6.0 ± 1.3 |
Statistical Test:
- p-value: 0.640 (not significant)
- Size reduction: 82%
Interpretation¶
All datasets show:
- No significant accuracy difference (p > 0.05)
- Substantial size reduction (46-82%)
- Maintained interpretability through small, simple trees
This validates the multi-objective approach: GA successfully balances accuracy and interpretability.