Table 4. Results of FS methods for SC-3.
| Feature selection | Pre process | Regressor | Wrapper | Number of features | Pearson | MSE |
|---|---|---|---|---|---|---|
| SC-3 Ph.1 | ||||||
| F Statis. | AF | Decision Tree R. | Gradient B. | 259 | 0.5990 | 0.2357 |
| Mutual Info.* | STPE | KNeighbors R. | ElasticNet | 121 | 0.5566 | 0.1893 |
| Mutual Info.* | STPE | KNeighbors R. | Gradient B. | 979 | 0.5443 | 0.1917 |
| F Statis. | STPE | Decision Tree R. | Lasso | 404 | 0.5346 | 0.1998 |
| SC-3 Ph.3 | ||||||
| F Statis.* | STPE | LinearSVR | ElasticNet | 1,736 | 0.6733 | 0.1920 |
| F Statis. | STPE | ElasticNet | ElasticNet | 1,736 | 0.6073 | 0.1973 |
| F Statis. | STPE | LinearSVR | ElasticNet | 1,736 | 0.6069 | 0.2062 |
| F Statis. | STPE | XGB Regressor | Lasso | 1,526 | 0.5693 | 0.2114 |
| Mutual Info.* | STPE | LinearSVR | Lasso | 4,410 | 0.5576 | 0.2021 |
| F Statis* | STPE | Ridge | ElasticNet | 1,736 | 0.5507 | 0.2148 |
| F Statis.* | AF | KNeighbors R. | ElasticNet | 671 | 0.5455 | 0.1990 |
Note:
Number of Features column represents number of distinct features selected. An asterisk (*) indicates that the hyper-parameters were not optimized.