Stress Classification Using Brain Signals Based on LSTM Network

Table 5

Performance metrics for stress classification using various classification techniques and training-testing set partitions.

Method	Validation method	Specificity	Recall	F1-score	Precision
MLP	50–50	50.17  4.97	79.06  2.35	73.17  3.35	67.66  4.66
LSTM 1	50–50	61.86  3.23	84.67  4.16	79.17  4.73	74.33  4.79
LSTM 2	50–50	88.04  4.21	86.81  4.84	90.04  2.11	93.53  3.23
MLP	60–40	51.19  5.63	82.90  3.43	73.83  4.43	65.14  4.34
LSTM 1	60–40	63.51  3.24	85.41  5.23	80.34  4.39	75.04  5.66
LSTM 2	60–40	86.69  4.44	90.66  4.86	91.68  3.91	82.72  4.27
MLP	70–30	60.60  4.65	79.01  2.28	64.51  3.25	77.26  6.32
LSTM 1	70–30	70.43  4.44	82.16  4.53	83.01  3.93	81.90  6.36
LSTM 2	70–30	84.85  3.72	93.51  3.18	92.45  2.68	91.42  5.23
MLP	10-fold cross validation	61.41  2.23	79.12  3.01	76.81  3.11	76.33  4.69
LSTM 1	10-fold cross validation	71.79  3.65	84.96  4.67	84.62  4.37	82.08  5.53
LSTM 2	10-fold cross validation	88.47  3.42	95.45  2.32	94.94  3.76	94.44  4.43