Skip to main content
. 2022 Jul 1;17(2):385–398. doi: 10.1007/s11571-022-09832-z

Table 3.

Metrics of the LSTM classifier with different time windows

Time window (s)
1 2 3 4 5 6 7 8 9 10
Accuracy
Mean 0.7661 0.8084 0.8401 0.8735 0.8787 0.8804 0.9094 0.9127 0.9065 0.9214
sd.s 0.1606 0.1496 0.1454 0.1367 0.1199 0.1310 0.1057 0.0911 0.1062 0.0935
Sensitivity
Mean 0.7075 0.7642 0.8139 0.8548 0.8563 0.8558 0.8990 0.8891 0.8967 0.9162
sd.s 0.1948 0.1971 0.1502 0.1721 0.1651 0.1741 0.1217 0.1147 0.1062 0.1267
Specificity
Mean 0.8247 0.8526 0.8662 0.8922 0.9011 0.9050 0.9199 0.9363 0.9163 0.9267
sd.s 0.1616 0.1518 0.1703 0.1153 0.1041 0.1268 0.1097 0.0802 0.1419 0.0640
Matthew correlation coefficient
Mean 0.5397 0.6316 0.6952 0.7563 0.7700 0.7776 0.8276 0.8373 0.8281 0.8534
sd.s 0.3268 0.2907 0.2815 0.2662 0.2279 0.2429 0.2014 0.1687 0.1939 0.1769

The sample mean and standard deviation are calculated as the average of 13 subjects with 20 runs each