Table 2.
Performance of ForSyn in different configurations under F1 score
| ForSyn(RFC∗2+ETFD∗2) | ForSyn(RFC∗3+ETFD∗3) | ForSyn(RFC∗4+ETFD∗4) | ForSyn(RFC∗4) | ForSyn(ETFD∗4) | |
|---|---|---|---|---|---|
| Data 1 | 0.499(2.5) | 0.491(4.0) | 0.499(2.5) | 0.341(5.0) | 0.510(1.0) |
| Data 2 | 0.496(3.0) | 0.525(1.0) | 0.501(2.0) | 0.364(5.0) | 0.477(4.0) |
| Data 3 | 0.519(3.0) | 0.543(1.5) | 0.543(1.5) | 0.327(5.0) | 0.460(4.0) |
| Data 4 | 0.529(1.0) | 0.524(2.0) | 0.519(3.0) | 0.349(5.0) | 0.475(4.0) |
| Data 5 | 0.568(3.0) | 0.575(1.5) | 0.575(1.5) | 0.335(5.0) | 0.497(4.0) |
| Data 6 | 0.551(1.5) | 0.539(3.0) | 0.551(1.5) | 0.345(5.0) | 0.473(4.0) |
| Data 7 | 0.564(1.5) | 0.564(1.5) | 0.547(3.0) | 0.354(5.0) | 0.493(4.0) |
| Data 8 | 0.572(1.0) | 0.556(2.0) | 0.547(4.0) | 0.339(5.0) | 0.551(3.0) |
| Average rank | 2.1 | 2.1 | 2.4 | 5.0 | 3.5 |
The value in parentheses represents the ranking value of the corresponding performance. Taking data 8 as an example, the ForSyn(RFC∗2+ETFD∗2) on this dataset has the best performance (0.572) and is assigned a ranking value of 1.0. In data 6, the performance of ForSyn(RFC∗2+ETFD∗2) and ForSyn(RFC∗4+ETFD∗4) are the same (0.551), and they occupy the first and second positions, respectively, so their ranking values are uniformly assigned 1.5 ([1.0 + 2.0]/2). The average rank of each algorithm is defined as the average of its ranks on all datasets. RFC, RF-CUS unit; ETFD, ETF-DR unit; and the number behind each unit represents the number of units of this type on each cascade layer. For example, ForSyn(RFC∗2+ETFD∗2) means that each cascade layer is placed with two RF-CUS units and two ETF-DR units.