Table 3.

Classification performance of the test dataset. Models include the constant (outcome bias), baseline models (k-nearest neighbors (KNN), logistic regression, naïve Bayes, decision tree), and black box models (support vector machine (SVM), neural network, AdaBoost, and random forest).

Activity (Resting, Web, Sudoku)	Liberal (95% Training)				Conservative (10% Training)
Model	AUC	F1	Precision	Recall	AUC	F1	Precision	Recall
Constant	0.49	0.22	0.16	0.40	0.50	0.22	0.16	0.39
K-Nearest Neighbors (k = 3)	0.97	0.93	0.93	0.93	0.89	0.84	0.84	0.84
Logistic Regression	0.52	0.23	0.44	0.41	0.36	0.28	0.41	0.41
Naïve Bayes	0.65	0.45	0.46	0.47	0.67	0.44	0.45	0.46
Decision Tree (depth = 4)	0.68	0.43	0.57	0.49	0.67	0.45	0.53	0.48
Support Vector Machine	0.50	0.31	0.33	0.32	0.44	0.28	0.31	0.29
Neural Network	0.79	0.61	0.61	0.61	0.33	0.45	0.47	0.47
AdaBoost	0.93	0.92	0.92	0.92	0.78	0.80	0.80	0.80
Random Forest	0.99	0.94	0.94	0.94	0.93	0.85	0.85	0.85