Skip to main content
. 2024 Dec 23;8:e63866. doi: 10.2196/63866

Table 5.

Prediction performance. Performance is reported in terms of prediction accuracy, precision, and recall for the MCIa category, and MCCb.

Classification Accuracy MCI precision MCI recall MCC
LLM

Demographics 0.48±0.00 0.47±0.00 0.33±0.00 –0.04±0.00

Behavior 0.43±0.02 0.43±0.02 0.43±0.04 –0.14±0.05

n-back 0.75±0.00 0.93±0.00 0.54±0.00 0.55±0.00

EMAc 0.65±0.00 0.89±0.00 0.33±0.00 0.37±0.00

Journal 0.57±0.04 0.56±0.03 0.63±0.05 0.13 ± 0.07

Test session 0.61±0.01 0.78±0.01 0.30±0.02 0.28±0.02

All 0.56±0.01 0.64±0.02 0.29±0.02 0.15±0.02

Wrapper 0.77±0.02 d 0.88±0.05 0.63±0.03 0.57± 0.05
Random forest 0.62±0.04 0.86±0.03 0.27±0.00 0.31±0.07
Gradient boosting 0.65±0.01 0.89±0.01 0.35±0.03 0.39± 0.02
XGBoost 0.67±0.00 0.83±0.00 0.42±0.00 0.38±0.00

aMCI: mild cognitive impairment.

bMCC: Matthews correlation coefficient.

cEMA: ecological momentary assessment.

dItalicized values indicate the best performer for each metric.