Table 3:
Mean (std.error) accuracy and true positive rates. View 1 data have signal variables with nonlinear relationships. TPR-1; true positive rate for X1. Deep IDA/Deep CCA/RKCCA view 1 means using the discriminant scores of view 1 only for classification. SVM view 1 uses view 1 data to train and test the model. – indicates not applicable
| Method | Mean (%) Accuracy | TPR-1 |
|---|---|---|
| Setting One | ||
| (p1 = 500, p2 = 500, n1 = 200, n2 = 150) | ||
| Deep IDA + Bootstrap | 60.87 (1.28) | 100.0 |
| Deep IDA + Bootstrap view 1 | 81.17 (2.89) | 100.0 |
| Deep IDA + TS | 62.60 (2.02) | 10.20 |
| Deep IDA + TS View 1 | 81.49 (3.06) | 10.20 |
| Deep CCA + TS | 58.20(0.59) | 8.30 |
| Deep CCA + TS view 1 | 61.26(0.77) | 8.30 |
| Deep CCA on top 50 selected features | 58.91(0.82) | - |
| Deep CCA on top 50 selected features view 1 | 59.87(0.87) | - |
| RKCCA | 61.06 (0.47) | - |
| RKCCA View 1 | 64.93 (0.63) | - |
| RKCCA on top 50 selected features | 56.21 (0.77) | - |
| RKCCA View 1 on top 50 selected features | 58.94 (0.78) | - |
| SVM | 54.20(0.30) | - |
| SVM view 1 | 50.26(0.21) | - |
| SVM on top 50 selected features | 50.37(0.27) | - |
| SVM on top 50 selected features view 1 | 50.07(0.26) | - |
| Setting Two | ||
| (p1 = 500, p2 = 500, n1 = 3, 000, n2 = 2, 250) | ||
| Deep IDA + Bootstrap | 89.45(2.16) | 63.70 |
| Deep IDA + Bootstrap View 1 | 90.49(2.25) | 63.70 |
| Deep IDA + TS | 91.78 (1.73) | 10.50 |
| Deep IDA + TS View 1 | 84.37 (1.49) | 10.50 |
| Deep CCA + TS | 59.84(0.40) | 33.70 |
| Deep CCA + TS view 1 | 60.35(0.34) | 33.70 |
| Deep CCA on top 50 selected features | 58.50(0.52) | - |
| Deep CCA on top 50 selected features view 1 | 58.32(0.52) | - |
| RKCCA | 57.14 (0.00) | - |
| RKCCA View 1 | 57.14 (0.00) | - |
| RKCCA on top 50 selected features | 66.30 (0.96) | - |
| RKCCA View 1 on top 50 selected features | 66.32 (0.98) | - |
| SVM | 54.42(0.11) | - |
| SVM view 1 | 52.81(0.13) | |
| SVM on top 50 selected features | 50.56(0.07) | - |
| SVM on top 50 selected features view 1 | 50.49(0.04) | - |
| Setting Three | ||
| (p1 = 2000, p2 = 2000, n1 = 200, n2 = 150) | ||
| Deep IDA + Bootstrap | 54.77(0.91) | 96.05 |
| Deep IDA + Bootstrap View 1 | 70.40(2.17) | 96.05 |
| Deep IDA + TS | 61.67 (1.74) | 30.55 |
| Deep IDA + TS View 1 | 60.73 (1.76) | 30.55 |
| Deep CCA + TS | 62.27(0.46) | 10.28 |
| Deep CCA + TS view 1 | 70.83(0.36) | 10.28 |
| Deep CCA on top 200 selected features | 61.43(0.62) | - |
| Deep CCA on top 200 selected features view 1 | 68.67(0.83) | - |
| RKCCA | 60.24 (0.63) | - |
| RKCCA View 1 | 63.54 (0.57) | - |
| RKCCA on top 200 selected features | 58.70 (0.52) | - |
| RKCCA View 1 on top 200 selected features | 61.90 (0.91) | - |
| SVM | 54.69(0.45) | - |
| SVM view 1 | 53.97(0.41) | - |
| SVM on top 200 selected features | 51.14(0.42) | - |
| SVM on top 200 selected features view 1 | 50.19(0.35) | - |
| Setting Four | ||
| (p1 = 2000, p2 = 2000, n1 = 3, 000, n2 = 2, 250) | ||
| Deep IDA + Bootstrap | 69.38(1.44) | 83.20 |
| Deep IDA + Bootstrap View 1 | 71.34(1.66) | 83.20 |
| Deep IDA + TS | 64.52 (1.32) | 10.78 |
| Deep IDA + TS View 1 | 69.99 (2.88) | 10.78 |
| Deep CCA + TS | 60.33(0.76) | 33.48 |
| Deep CCA + TS view 1 | 58.54(1.00) | 33.48 |
| Deep CCA on top 200 selected features | 59.58(1.66) | - |
| Deep CCA on top 200 selected features view 1 | 59.28(1.71) | - |
| RKCCA | 57.14 (0.00) | - |
| RKCCA View 1 | 57.14 (0.00) | - |
| RKCCA on top 200 selected features | 55.30 (0.79) | - |
| RKCCA View 1 on top 200 selected features | 61.21 (0.81) | - |
| SVM | 53.13(0.16) | - |
| SVM view 1 | 53.37(0.13) | - |
| SVM on top 200 selected features | 52.72(0.13) | - |
| SVM on top 200 selected features view 1 | 50.59(0.12) | - |