Research Article
Zheng Classification with Missing Feature Values Using Local-Validity Approach
Table 4
The performance comparison of three methods on three public datasets.
(a) Lymphography |
| | Diagnosis accuracy (%) | Deletion | Imputation | Local-validity |
| 0 | 85.14 | 85.14 | 83.7 | 0.05 | 85.82 | 83.78 | 84.02 | 0.10 | 81.21 | 81.08 | 92.16 | 0.20 | 78.92 | 77.43 | 88.11 |
|
|
(b) SPECT heart |
| | Diagnosis accuracy (%) | Deletion | Imputation | Local-validity |
| 0 | 82.4 | 82.40 | 82.05 | 0.05 | 82.28 | 82.02 | 83.06 | 0.10 | 82 | 80.52 | 85.25 | 0.20 | 80.2 | 78.08 | 86.06 |
|
|
(c) Lung cancer |
| | Diagnosis accuracy (%) | Deletion | Imputation | Local-validity |
| 0 | 90.6 | 90.6 | 90.05 | 0.05 | 89.25 | 90.12 | 89.17 | 0.10 | 86.23 | 83.37 | 87.29 | 0.20 | 81.67 | 82.02 | 82.37 |
|
|
The bold values are used to emphasize the best Zheng classification performance.
|