Many different rule interestingness measures have been proposed in the literature; we show that, under two assumptions, at least twelve of these measures are proportional to Confidence. We consider rules with a fixed consequent, generated from a fixed data set. From these assumptions, we prove that Satisfaction, Ohsaki's Conviction, Added Value, Brin's Interest/Lift/ Strength, Brin's Conviction, Certainty Factor/Loevinger, Mutual Information, Interestingness, Sebag-Schonauer, Ganascia Index, Odd Multiplier, and Example/counter-example Rate are all monotonic with respect to Confidence. Hence, for ordering sets of partial classification rules with a fixed consequent, the Confidence measure is equivalent to any of the twelve other measures. © 2012 Springer-Verlag.
CITATION STYLE
Hills, J., Davis, L. M., & Bagnall, A. (2012). Interestingness measures for fixed consequent rules. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7435 LNCS, pp. 68–75). https://doi.org/10.1007/978-3-642-32639-4_9
Mendeley helps you to discover research relevant for your work.