Skip to main content
. Author manuscript; available in PMC: 2021 May 9.
Published in final edited form as: Acta Neurol Scand. 2020 Jan 22;141(5):388–396. doi: 10.1111/ane.13216

TABLE 4.

Classifier performance for identifying individual disorders. The number of features used for classification was a parameter tuned separately from the other hyperparameters. The AROC is the Area under the Receiver Operating Characteristic curve. The P-values represent the significance of the AROCs compared to a completely random classifier. (a) Leave-one-out (LOO) cross-validation performance using the first interviews with all participants. The feature selection for the cross-validation was performed within each fold; however, the “top features” presented here use the same method applied to the entire dataset. (b) Training sets constructed from stratified random samples of 50% of first interviews and 50% of second interviews with performance tested against all remaining interviews

(a)
Suicidality (n = 30/122) Despressive Disorders s (n = 29/122) Anxiety Disorders (n = 15/122) Bipolar Disorders (n = 6/122)
AROC (95% CI) 57 (44-71)% 65 (54-76)% 66 (52-80)% 78 (64-92)%
P-value .22 .011 .04 .003
Total No. Features 16 512 32 64
Top 5 features for presence of comorbidity (SVM weights), try to, sometimes, things, I do not really, my its, is, yes, like i, hope I feel, yes, myself, my, try to it is, be, weird, now, things
Top 5 features for absence of comorbidity (SVM weights) I have, in, try, assent (LIWC), inhibit (LIWC) fears, if i, on, get, know no, and I, to, that I, feel get, angry, for, think, so
(b)
Suicidality (n = 26/114) Depressive Disorders (n = 25/113) Anxiety Disorders (n = 12/112) Bipolar Disorders (n = 4/114)
AROC (95% CI) 71 (59-83)% 69 (58-81)% 64 (46-81)% 66 (34-99)%
P-value .0011 .001 .12 .3
Total No. Features 64 128 256 32
 Top 5 features for presence of comorbidity (SVM weights) like a, fear, job, things, always uh, like that, in, that i, was kinda, going, no, more,
you
time, i just, or, pretty,
thats
 Top 5 features for absence of comorbidity (SVM weights) wrong, a lot, makes me, feel like, really i am, she, at, talk to, come myself, most, lot of, a lot of, of secrets, give, anything to, to explain, thats really