Premium
P1‐454: VOCALIC MARKERS OF COGNITIVE LOAD DERIVED FROM AUTOMATED VERBAL NEUROPSYCHOLOGICAL ASSESSMENT AND MACHINE LEARNING IN A LARGE SCALE REMOTE SAMPLE
Author(s) -
Taptiklis Nick,
Su Merina,
Barnett Jennifer H.,
Cormack Francesca K.
Publication year - 2019
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2019.06.1059
Subject(s) - naive bayes classifier , random forest , memory span , context (archaeology) , cognition , support vector machine , neuropsychology , task (project management) , computer science , psychology , discriminative model , artificial intelligence , machine learning , speech recognition , working memory , paleontology , management , neuroscience , economics , biology
10-fold repeated cross-validation resampling method using caret R package on the train dataset. The model with the highest predicted accuracy on test data was kept. Finally accuracy and other diagnostic performance variables of final models were calculated using the test dataset. Results: Complete data was available from 607 patients. No CI, MCI or dementia was diagnosed in 164, 363 and 80 subjects respectively.Random forest models showed the highest predicted test accuracy both for CI presence and type classification. Accuracy on test data set were 0.88 (0.82-0.92) and 0.91 (0.840.95) respectevily. High positive and negative predicted values were observed for CI and dementia prediction (table 1). Conclusions: In subjects with subjective cognitive complaints, a sequential classification using a random forest machine learning algorith may be clinically useful.