Table 4 Feature selection performance of different approaches in simulated scenarios for interaction models
From: Artificial Intelligence based wrapper for high dimensional feature selection
Methods | Performance (Number of Features Selected) | |||||
|---|---|---|---|---|---|---|
Interaction Model Scenarios | ||||||
1_M | 2_M | 3_M | ||||
p = 15 | χ = 105 | p = 25 | χ = 300 | p = 50 | χ = 1225 | |
Target Features | ||||||
10 | 9 | 10 | 9 | 10 | 9 | |
Mean (Range) | ||||||
ALASSO | 15 (15–15) | 31 (20–41) | 24 (22–25) | 46 (32–67) | 32 (2–45) | 36 (1–67) |
GLASSO | 15 (14–15) | 40 (22–51) | 25 (24–25) | 66 (39–74) | 47 (45–49) | 76 (72–81) |
LASSO | 15 (15–15) | 33 (18–49) | 24 (22–25) | 45 (30–65) | 16 (1–45) | 16 (0–71) |
SPLS | 14 (12–15) | 36 (16–102) | 19 (9–25) | 65 (6–287) | 38 (6–50) | 417 (1–1057) |
Enet | 15 (15–15) | 34 (21–44) | 22 (14–25) | 39 (11–60) | 29 (2–50) | 36 (1–116) |
AEnet | 15 (15–15) | 32 (24–41) | 24 (22–25) | 44 (31–64) | 37 (2–49) | 53 (1–104) |
AIWRAP-L | 12 (12–14) | 34 (20–47) | 18 (14–21) | 50 (26–60) | 29 (27–32) | 85 (71–99) |
AIWRAP-LLr | 12 (12–14) | 30 (8–44) | 16 (10–20) | 36 (5–47) | 24 (8–30) | 30 (2–52) |
AIWRAP-LR | 12 (12–14) | 34 (20–47) | 18 (14–21) | 50 (26–60) | 28 (24–30) | 46 (26–88) |