Research Article
IGPred-HDnet: Prediction of Immunoglobulin Proteins Using Graphical Features and the Hierarchal Deep Learning-Based Approach
Table 1
Analysis of various classifiers using feature encoding schemes on training and testing datasets Dtrain and Dtest.
| Feature-encoding methods | Classifiers | Benchmark dataset | Independent dataset | ACC (%) | SN (%) | SP (%) | MCC | F-measure (%) | ACC (%) | SN (%) | SP (%) | MCC | F-measure (%) |
| APAAC | KNN | 95.01 | 94.55 | 95.26 | 0.898 | 92.95 | 88.93 | 97.50 | 83.33 | 0.777 | 85.71 | DT | 90.70 | 87.27 | 92.66 | 0.802 | 86.92 | 91.96 | 92.50 | 91.66 | 0.829 | 89.15 | SVM | 89.72 | 86.36 | 91.66 | 0.786 | 85.61 | 81.25 | 82.50 | 80.55 | 0.612 | 75.86 | HDnet | 95.69 | 91.82 | 93.75 | 0.909 | 93.83 | 90.17 | 87.50 | 91.66 | 0.787 | 86.41 |
| DPC | KNN | 90.41 | 91.82 | 89.63 | 0.809 | 86.77 | 92.85 | 97.50 | 90.27 | 0.854 | 90.69 | DT | 89.39 | 85.45 | 91.66 | 0.771 | 85.11 | 93.75 | 92.05 | 94.44 | 0.864 | 91.35 | SVM | 90.23 | 96.36 | 45.58 | 0.467 | 69.67 | 83.03 | 100 | 73.61 | 0.706 | 80.80 | HDnet | 96.02 | 90.00 | 99.47 | 0.916 | 94.18 | 91.96 | 77.50 | 100 | 0.829 | 87.32 |
| FEGS | KNN | 93.03 | 90.00 | 94.78 | 0.856 | 89.71 | 94.64 | 100 | 91.60 | 0.890 | 93.00 | DT | 91.93 | 87.27 | 93.74 | 0.817 | 87.95 | 93.75 | 97.50 | 91.60 | 0.871 | 91.76 | SVM | 94.72 | 94.55 | 94.85 | 0.889 | 92.96 | 93.75 | 100 | 90.27 | 0.876 | 91.95 | HDnet | 98.00 | 94.55 | 100 | 0.958 | 96.94 | 99.10 | 97.50 | 100 | 0.980 | 98.73 |
|
|