Skip to main content
. 2016 Jul 12;77(6):2114–2124. doi: 10.1002/mrm.26318

Table 3.

BAR of the pattern Recognition Techniques Obtained Using Metabolite Profiles and Complete Spectra as the Classifier Input at all bSMOTE Oversampling Rates.

Original 0.5 1 1.5 2
AdaBoostM1‐NB
Spectra 0.56 0.82 0.80 0.82 0.80
Metabolite 0.70 0.76 0.82 0.90 0.89
AdaBoostM1‐SVM
Spectra 0.67 0.83 0.87 0.90 0.86
Metabolite 0.76 0.78 0.88 0.90 0.93
AdaBoostM1‐NN
Spectra 0.63 0.84 0.79 0.83 0.82
Metabolite 0.78 0.83 0.93 0.92 0.92
AdaBoostM1‐LDA
Spectra 0.58 0.70 0.76 0.80 0.80
Metabolite 0.82 0.79 0.86 0.93 0.91
Random forests
Spectra 0.73 0.79 0.75 0.84 0.80
Metabolite 0.72 0.76 0.82 0.85 0.90