Introduction
Overview of complex-step perturbation approach (CSPA)
Complex-step feature selection method
Feature selection for regression using complex-step sensitivity
Feature selection for classification using complex-step sensitivity
Numerical experiments
Datasets
Regression
Classification
Dataset name | Instances | No. of features | No. of target variables |
---|---|---|---|
Bodyfat | 252 | 13 | 1 |
Abalone | 4177 | 10 | 1 |
Wine quality | 4898 | 11 | 1 |
Dataset name | Instances | No. of features | No. of class labels |
---|---|---|---|
Vehicle | 846 | 30 | 4 |
Segmentation | 210 | 18 | 7 |
Breast cancer | 569 | 18 | 2 |
Configuring feed-forward neural networks
Results
Regression
Bodyfat dataset | Abalone dataset | Wine quality dataset | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Corr | ReliefF | MI | CS-FS | Corr | ReliefF | MI | CS-FS | Corr | ReliefF | MI | CS-FS |
6 | 6 | 6 | 6 | 10 | 10 | 10 | 7 | 11 | 2 | 8 | 11 |
5 | 5 | 5 | 3 | 5 | 7 | 5 | 8 | 9 | 11 | 11 | 4 |
7 | 7 | 7 | 13 | 6 | 8 | 7 | 6 | 10 | 6 | 4 | 6 |
2 | 2 | 2 | 4 | 4 | 9 | 6 | 10 | 6 | 9 | 7 | 2 |
8 | 8 | 8 | 8 | 7 | 5 | 9 | 9 | 3 | 7 | 5 | 7 |
9 | 9 | 9 | 2 | 9 | 6 | 4 | 4 | 4 | 1 | 6 | 5 |
1 | 1 | 11 | 1 | 8 | 4 | 8 | 5 | 1 | 10 | 3 | 1 |
11 | 11 | 4 | 7 | 1 | 2 | 2 | 2 | 7 | 8 | 2 | 9 |
3 | 3 | 1 | 5 | 3 | 3 | 1 | 3 | 2 | 3 | 9 | 3 |
4 | 4 | 13 | 12 | 2 | 1 | 3 | 1 | 5 | 4 | 1 | 8 |
10 | 10 | 12 | 11 | 8 | 5 | 10 | 10 | ||||
13 | 13 | 10 | 10 | ||||||||
12 | 12 | 3 | 9 |
Classification
Method | Feature ranking |
---|---|
Vehicle dataset | |
ReliefF | 8, 7, 12, 9, 3, 11, 18, 4, 2, 1, 13, 10, 16, 14, 17, 6, 15, 5 |
Symmetric uncertainty | 12, 7, 8, 11, 9, 6, 3, 4, 1, 13, 2, 14, 10, 17, 18, 5, 16, 15 |
Info gain | 12, 7, 8, 11, 9, 3, 6, 2, 1, 4, 13, 10, 14, 17, 18, 5, 16, 15 |
Gain ratio | 11, 9, 12, 7, 4, 8, 6, 3, 5, 18, 13, 14, 1, 2, 16, 10, 15, 17 |
Chi-squared | 12, 7, 8, 9, 11, 3, 6, 1, 2, 10, 14, 13, 4, 17, 18, 5, 16, 15 |
CS-FS | 10, 8, 5, 17, 14, 18, 11, 3, 6, 12, 7, 1, 9, 4, 13, 2, 15, 16 |
Segmentation dataset | |
ReliefF | 11, 16, 18, 9, 12, 10, 2, 15, 14, 13, 17, 1, 5, 7, 3, 4, 6, 8 |
Symmetric uncertainty | 18, 10, 9, 16, 12, 11, 15, 17, 2, 14, 13, 7, 8, 5, 6, 3, 4, 1 |
Info gain | 18, 9, 12, 16, 10, 11, 15, 17, 13, 14, 2, 7, 8, 5, 6, 3, 4, 1 |
Gain ratio | 10, 11, 9, 16, 18, 2, 12, 14, 15, 17, 13, 8, 7, 5, 6, 3, 4, 1 |
Chi-squared | 18, 12, 9, 16, 10, 11, 13, 15, 17, 14, 2, 7, 8, 5, 6, 3, 4, 1 |
CS-FS | 2, 18, 15, 13, 10, 16, 11, 12, 17, 9, 14, 6, 8, 7, 5, 4, 3, 1 |
Breast cancer dataset | |
ReliefF | 28, 8, 21, 23, 3, 1, 7, 24, 4, 27, 26, 6, 22, 25, 11, 2, 14, 13, 29, 30, 10, 18, 5, 16, 9, 17, 19, 15, 12, 20 |
Symmetric uncertainty | 23, 21, 24, 28, 8, 3, 7, 4, 1, 27, 14, 11, 13, 6, 26, 17, 2, 18, 22, 25, 29, 16, 5, 30, 9, 19, 20, 10, 12, 15 |
Info gain | 23, 24, 21, 28, 8, 3, 4, 1, 7, 14, 27, 11, 13, 26, 6, 17, 18, 22, 2, 29, 16, 25, 9, 5, 30, 20, 19, 10, 12, 15 |
Gain ratio | 23, 21, 24, 28, 8, 7, 27, 3, 4, 1, 14, 6, 11, 13, 26, 17, 2, 19, 18, 25, 22, 29, 5, 16, 30, 9, 20, 12, 10, 15 |
Chi-squared | 23, 21, 24, 28, 8, 3, 4, 1, 7, 14, 27, 11, 13, 26, 6, 17, 18, 22, 2, 29, 25, 16, 9, 5, 30, 20, 19, 10, 12, 15 |
CS-FS | 21, 23, 28, 20, 8, 4, 7, 11, 24, 17, 15, 2, 22, 30, 12, 26, 13, 16, 1, 14, 10, 9, 29, 25, 18, 19, 6, 3, 27, 5 |