Research Article

A Novel Approach to Ensemble Classifiers: FsBoost-Based Subspace Method

Table 12

Comparing the FsBoost algorithm with single classifiers and boosting algorithms for other datasets.

MethodDatasetsMean
BasehockMadelonPCMAC
RankAccRankAccRankAccAccRank

AdaBoostM1559.341752.92263.7558.675
Bag1750.90855.001752.7352.8817
GentleBoost1358.231553.54562.8258.209
LogitBoost1258.431453.69363.5458.566
LPBoost1656.22755.151652.8354.7416
RobustBoost459.541353.77163.9559.092
RUSBoost1556.33656.001256.0256.1214
Subspace1158.551653.311554.0455.3015
TotalBoost659.041254.38462.8258.754

FsBoost.V1
Level 1 kNN ensemble1058.73356.851454.2756.6212
Level 1 PNN ensemble259.74256.921158.8158.498
Level 1 SVM ensemble758.841154.541060.1457.8411
Level 2 ensemble160.14456.62960.5659.101

FsBoost.V2
Level 1—ensemble 1958.841054.62662.2058.557
Level 1—ensemble 2858.84954.69760.8758.1310
Level 1—ensemble 31457.33157.381354.8956.5413
Level 2 ensemble359.74556.54860.7659.013
kNN59.0454.2354.58
PNN57.2353.0058.39
SVMs56.2254.6961.89

Acc: accuracy (%).