Towards automatic text-based estimation of depression through symptom prediction

Brain Informatics

Table 4 Test scores for each symptom

Symptom	MAE \(\pm \sigma\)	RRMSE \(\pm \sigma\)	miF1 \(\pm \sigma\)	maF1 \(\pm \sigma\)
No interest	0.529 ± 0.047	0.877 ± 0.067	0.800 ± 0.024	0.669 ± 0.043
Depressed mood	0.550 ± 0.027	0.733 ± 0.022	0.821 ± 0.019	0.729 ± 0.024
Insomnia or hypersomnia	0.753 ± 0.073	0.805 ± 0.060	0.774 ± 0.055	0.757 ± 0.047
Feeling tired	0.638 ± 0.031	0.816 ± 0.030	0.745 ± 0.030	0.709 ± 0.035
Eating too little or too much	0.811 ± 0.049	0.972 ± 0.064	0.762 ± 0.035	0.685 ± 0.026
Feeling of being a failure	0.620 ± 0.018	0.796 ± 0.012	0.817 ± 0.024	0.779 ± 0.021
Problems with concentrating	0.830 ± 0.040	0.878 ± 0.012	0.681 ± 0.034	0.557 ± 0.029
Moving too slowly or too fast	0.438 ± 0.022	0.976 ± 0.035	0.936 ± 0.000	0.484 ± 0.000

All models were run five times with different seed values, and the average values with standard deviation are presented. For computing the F1 scores, the predicted scores were binarized, such that the scores \(< 1.5\) were treated as negative class instances, and the scores \(\ge 1.5\) were treated as positive class instances