Skip to main content
. 2021 Apr 12;18(8):4069. doi: 10.3390/ijerph18084069

Table 1.

Performance of the Bi-LSTM-128 models on the development set.

Learning Rate Epoch Accuracy Precision Recall F1 Score AUC
0.00003 10 77.2% 26.2% 74.9% 38.9% 84.2%
20 78.7% 28.1% 76.9% 41.1% 85.1%
30 79.8% 28.7% 73.3% 41.2% 86.3%
40 82.3% 31.2% 68.6% 42.9% 87.1%
50 82.3% 31.3% 69.3% 43.2% 87.6%
60 82.7% 31.8% 68.6% 43.4% 87.7%
70 81.3% 30.1% 70.6% 42.2% 87.6%
80 82.5% 31.9% 71.3% 44.0% 88.0%
0.0001 10 80.7% 29.9% 73.9% 42.6% 87.2%
20 80.5% 29.9% 75.6% 42.8% 88.3%
30 80.8% 30.1% 74.9% 43.0% 88.0%
40 85.8% 37.0% 66.3% 47.5% 88.2%
50 89.7% 47.2% 53.1% 50.0% 86.2%
60 88.4% 43.2% 64.4% 51.7% 87.9%
70 86.1% 37.4% 64.0% 47.2% 87.3%
80 88.8% 43.9% 54.1% 48.4% 85.0%
0.001 10 84.5% 35.2% 71.3% 47.1% 88.1%
20 84.7% 35.1% 68.0% 46.3% 86.9%
30 90.0% 47.9% 41.6% 44.5% 83.8%
40 89.1% 44.6% 52.1% 48.1% 80.4%
50 90.7% 53.3% 34.3% 41.8% 74.5%
60 89.6% 46.2% 42.6% 44.3% 78.3%
70 88.6% 42.1% 47.2% 44.5% 79.4%
80 88.7% 42.5% 46.5% 44.4% 77.5%