Multilabel Classification Using Low-Rank Decomposition

Figure 1

Comparison of LLRD (control algorithm) against other related approaches with the Bonferroni–Dunn test. Approaches that are not connected to LLRD are significantly different in performance from LLRD. (a) One-error. (b) Coverage. (c) Ranking loss. (d) Average precision. (e) Macroaveraging F1. (f) Microaveraging F2.

(a)

(b)

(c)

(d)

(e)

(f)