Research Article

Novel Feature Selection Method for Nonlinear Support Vector Regression

Table 5

Comparison results of FS-NSVR, L1-SVR, Lp-SVR, and L1-LSSVR for real-world datasets.

Data setsRegressor (%)NMSERMSECPU S.

StudentFS-NSVR6.2500.0940.0370.9130.070
L1-SVR72.9170.0620.0300.9561.118
Lp-SVR62.5000.0650.0300.9800.400
L1-LSSVR52.0830.0620.0250.9600.001

TrafficFS-NSVR18.5190.9050.9460.1022.007
L1-SVR85.1850.9630.9760.2522.259
Lp-SVR88.8891.0831.0350.5300.457
L1-LSSVR44.4440.9521.6020.0570.001

CNNFS-NSVR7.4070.0010.0351.0000.237
L1-SVR86.4200.000.0111.0001.249
Lp-SVR9.8770.0010.0320.9810.372
L1-LSSVR8.6420.0000.0200.9830.001

CommunityFS-NSVR21.5690.3790.1380.9181.108
L1-SVR66.6670.3310.1290.6562.047
Lp-SVR37.2550.3370.1300.6371.822
L1-LSSVR74.5100.3400.1300.7310.001

PakinsonFS-NSVR23.0770.8620.4460.4151.176
L1-SVR42.3081.0930.5020.3191.276
Lp-SVR88.4621.2290.5330.4300.570
L1-LSSVR76.9231.0210.3720.1970.001