Table 7.

Confusion matrix for multi-class classification.

Result	Best-Submission				Selection-Submission
	ad	hc	mci	cmci	ad	hc	mci	cmci
AD	35	0	0	5	37	0	0	3
HC	1	15	6	18	0	18	12	10
MCI	7	8	5	20	3	15	11	11
cMCI	8	3	0	29	4	8	7	21
Precision	0.69	0.58	0.45	0.40	0.84	0.44	0.37	0.47
Recall	0.88	0.38	0.13	0.73	0.93	0.45	0.28	0.53
F1-score	0.77	0.46	0.20	0.52	0.88	0.44	0.31	0.49
Official Score	52.500%				54.387%

The lowercase ad, hc, mci, and cmci denotes the number of AD, HC, MCI and cMCI which generated from our algorithm. The capital AD, HC, MCI, and cMCI mean the number of real data (all these numbers exclude the simulated data).