RQ4 : Motifs cardinality

RQ4.1

Are the methods capable of finding all occurrences of the most relevant motifs independently of their cardinality?

In this section, we evaluate the ability of methods to detect the occurrence of motifs with different cardinalities. Our evaluation is summarized in the plot below.

Summary of the results

RQ4.1 Conclusion

VALMOD, PanMP, STOMP, Grammarviz, LatentMotif and SetFinder are particularly robust to the number of occurrences. Motiflets has good performances when motifs sets have many occurences.

RQ4.2

Are the methods capable of detecting several motifs when there is a cardinality unbalance?

In this section, we evaluate the impact of an unbalanced cardinality between different motifs, i.e., a difference in the number of occurrences between two different motifs. Our evaluation is summarized in the figure below, which depicts the average f1score on 100 runs for several methods . The diagonal corresponds to cases where the two motifs have the same number of occurrences.

Summary of the results

RQ4.2 Conclusion

VALMOD, STOMP, and PanMP are particularly robust to unbalanced cardinality of motif sets.