Research Article
Feature Selection and Parameter Optimization of Support Vector Machines Based on Modified Artificial Fish Swarm Algorithms
Table 5
Result of Experiment 1.
|
Dataset | Number of original features | Group 1 | Group 2 | Group 3 | No feature selection | Feature selection by AFSA | Feature selection by MAFSA | Number of selected features | Average accuracy rate (%) | Number of selected features | Average accuracy rate (%) | Number of selected features | Average accuracy rate (%) |
| Bupa | 6 | 6 | 58.85 | 3.3 | 59.73** | 3.2* | 59.73** | Pima | 8 | 8 | 77.21 | 5* | 80.47** | 5.6 | 80.47** | Glass | 9 | 9 | 48.11 | 4.8 | 64.89** | 4.4* | 64.89** | Vowel | 10 | 10 | 71.91 | 9 | 77.57 | 8.5* | 77.67** | Heart | 13 | 13 | 81.85 | 7.2 | 94.07** | 7.1* | 94.07** | Australian | 14 | 14 | 85.79 | 8 | 88.26** | 7.7* | 88.26** | Vehicle | 18 | 18 | 71.98 | 12.4* | 75.41 | 12.4* | 75.53** | Robot | 24 | 24 | 85.90 | 17* | 86.73 | 17.6 | 87.48** | German | 24 | 24 | 74.90 | 14.7 | 82.90 | 12.4* | 83.50** | Sonar | 60 | 60 | 73.19 | 29.7 | 93.33 | 27.2* | 94.28** |
|
|
It indicates the fewest number of selected features. It indicates the highest classification accuracy.
|