Table 3. Metrics about psychometric predicted validity of the Episodix, semantic memory and procedural memory games.
EPISODIX+S+PDATASET | Recall | Precision | ||||||
---|---|---|---|---|---|---|---|---|
ML classifier | F1 | k ↓ | HC | MCI | AD | HC | MCI | AD |
ET: Extra Trees | 0,97 | 0,97 | 0,96 | 0,99 | 0,99 | 0,99 | 0,91 | 1,00 |
RF: Random Forest | 0,97 | 0,96 | 0,95 | 0,98 | 0,99 | 0,98 | 0,91 | 1,00 |
GB: Gradient Boosting | 0,96 | 0,95 | 0,95 | 0,94 | 0,98 | 0,96 | 0,93 | 0,98 |
SVM: Support Vector Machine | 0,91 | 0,90 | 0,92 | 0,82 | 0,99 | 0,90 | 0,87 | 0,98 |
LR: Logistic Regression | 0,84 | 0,81 | 0,77 | 0,80 | 0,98 | 0,90 | 0,67 | 0,95 |
AB: Ada Boost | 0,84 | 0,80 | 0,77 | 0,78 | 0,99 | 0,90 | 0,70 | 0,93 |
Notes.
Episodix, semantic memory and procedural memory dataset (458 triples). ML algorithms: GB, Gradient Boosting classifier; ET, Extra Trees classifier; SVM, Support Vector Machines; LR, Logistic regression; AB, Ada Boost classifier; and RF, Random forest. Metrics: F1 score; Cohen’s Kappa (i.e., used as index to order best classification); recall and precision, the last two distributed by cognitive group. Experiments were performances with cv-fold cross validation (cv = 55) and default configuration in ML algorithms.