Skip to main content

Table 3 Average classification accuracy (± standard deviation) of common feature set

From: A machine learning approach to predict perceptual decisions: an insight into face pareidolia

Subject

Classification performance of individual subjects (in %)

TFPS39

TFPSL17

TFPSR17

DATFPS17

p value: 0.035

p value: 0.03

p value: 0.035

p value: 0.045

Subject1

\(69.98 \pm 6.11\)

\(66.92 \pm 6.70\)

\(63.80 \pm 6.19\)

\(\mathit{70 }.\mathit{05 } \pm \mathit{7 }.\mathit{16 }\)

Subject2

\(73.08 \pm 6.27\)

\(70.38 \pm 6.73\)

\(67.74 \pm 6.84\)

\(\mathit{73 }.\mathit{30 } \pm \mathit{6 }.\mathit{26 }\)

Subject3

\(\mathit{70 }.\mathit{67 } \pm \mathit{6 }.\mathit{57 }\)

\(65.81 \pm 6.56\)

\(62.69 \pm 5.45\)

\(69.74 \pm 6.76\)

Subject4

\(72.67 \pm 5.72\)

\(69.43 \pm 7.13\)

\(65.62 \pm 6.30\)

\(\mathit{74 }.\mathit{33 } \pm \mathit{6 }.\mathit{52 }\)

Subject5

\(69.86 \pm 6.83\)

\(65.58 \pm 6.84\)

\(68.17 \pm 6.76\)

\(\mathit{71 }.\mathit{03 } \pm \mathit{7 }.\mathit{16 }\)

Subject6

\(\mathit{75 }.\mathit{04 } \pm \mathit{5 }.\mathit{83 }\)

\(67.86 \pm 6.41\)

\(69.43 \pm 6.93\)

\(72.82 \pm 6.38\)

Subject7

\(71.93 \pm 6.24\)

\(65.92 \pm 6.07\)

\(68.17 \pm 6.34\)

\(\mathit{72 }.\mathit{92 } \pm \mathit{6 }.\mathit{56 }\)

PAM

\(71.89 \pm 1.88\)

\(67.41 \pm 1.89\)

\(66.52 \pm 2.53\)

\(\mathit{72 }.\mathit{03 } \pm \mathit{1 }.\mathit{76 }\)

  1. PAM personalized average model, TFPS39 time–frequency power spectrum of 39 electrodes from common feature set (p < 0.035), TFPSL17 time–frequency power spectrum of 17 electrodes from left hemisphere (p < 0.03), TFPSR17 time–frequency power spectrum of 17 electrodes from right hemisphere (p < 0.035), DATFPS17 differential asymmetry of TFPS of 17 electrode pairs (p < 0.045). These p values are uncorrected
  2. For each subject, among four feature types, which yields highest performance are represented in italic form