Table 3.

The mean sensitivity and specificity scores with corresponding standard deviations for each class individually on the migbase dataset for the original training set and the transformed datasets obtained using three over-sampling techniques

Technique	Migraine		Tension		Cluster
	Sensitivity	Specificity	Sensitivity	Specificity	Sensitivity	Specificity
Prior knowledge	0.9848±0.003+	0.9753±0.006	0.9682±0.007	0 . 9 8 7 5 ± 0 . 0 0 3 +	0.9775±0.011+	0.9973±0.002
ADASYN	0.9839±0.003	0.9771±0.005+	0.9683±0.007−	0.9836±0.003−	0.9421±0.022−	0.9969±0.002
SMOTE	0.9845±0.003+	0.9767±0.006+	0.9721±0.008+	0.9845±0.003	0.9307±0.024−	0.9967±0.002
Sample weight	0.9830±0.003	0.9742±0.007	0.9696±0.008	0.9827±0.003−	0.9250±0.024−	0.9969±0.002
None	0.9834±0.003	0.9744±0.006	0.9695±0.008	0.9850±0.003	0.9556±0.021	0 . 9 9 7 4 ± 0 . 0 0 2

A cell is marked as.⁺ or.⁻ if the result is a statistically significant (ρ≤0.05) improvement or detriment respectively compared to the baseline (None), according to a bootstrap test