Table 4.
Average misclassification error rate and area under the receiver operating characteristic curve (AUC) of sample classification for two class real dataset (hepatocellular carcinoma) with 26.52% missing values and artificially imputed different rates of outliers.
| Methods | Without outliers MER (AUC) | 3% outliers MER (AUC) | 5% outliers MER (AUC) | 7% outliers MER (AUC) | 10% outliers MER (AUC) |
|---|---|---|---|---|---|
| RF | 10.67 (0.8903) | 13.61 (0.8642) | 20.16 (0.8001) | 21.18 (0.7883) | 22.61 (0.7751) |
| PPCA | 15.22 (0.8495) | 16.45 (0.8365) | 26.44 (0.7364) | 26.56 (0.7355) | 26.67 (0.7323) |
| kNN | 13.53 (0.8795) | 13.94 (0.8676) | 25.56 (0.7484) | 25.97 (0.7402) | 26.33 (0.7375) |
| BPCA | 14.61 (0.8571) | 16.28 (0.8342) | 21.67 (0.7882) | 23.38 (0.7679) | 25.45 (0.7452) |
| EM-PCA | 11.44 (0.8894) | 11.58 (0.8813) | 23.72 (0.7665) | 23.70 (0.7648) | 23.69 (0.7618) |
| Zero | 11.55 (0.8874) | 12.39 (0.8747) | 15.58 (0.8433) | 17.51 (0.8246) | 19.44 (0.8026) |
| Mean | 10.56 (0.8914) | 13.61 (0.8644) | 20.73 (0.7947) | 22.69 (0.7723) | 24.89 (0.7502) |
| Median | 9.94 (0.9068) | 12.38 (0.8783) | 17.61 (0.8276) | 20.17 (0.7975) | 23.54 (0.7637) |
| Minimum | 10.56 (0.8937) | 11.24 (0.8894) | 13.57 (0.8646) | 15.67 (0.8427) | 17.44 (0.8265) |
| rmiMAE | 0.78 (0.9922) | 1.84 (0.9815) | 2.36 (0.9773) | 2.97 (0.9701) | 3.65 (0.9644) |
| Proposed | 0.00 (1.00) | 0.32 (0.9972) | 0.75 (0.9929) | 1.18 (0.9895) | 1.87 (0.9837) |
Bold indicates the lower MER and Higher AUC throughout the column.