Data mining of serial-episode association rules using gap-constrained minimal occurrences
H.K. Dai, Z. Wang · International Journal of Business Intelligence and Data Mining · 2013
Data mining is a task of extracting useful patterns/episodes from large databases. Sequence data can be modelled using episodes. An episode is serial if the underlying temporal order is total. An episode rule of associating two episodes suggests a temporal implication of the antecedent episode to the consequent episode. We present two sound and complete mining algorithms for finding frequent and confident serial-episode association rules with their ideal occurrence/window widths, if existing, in event sequences based on the notion of minimal occurrences constrained by constant and mean maximum gap, respectively. Two empirical studies are summarised: one focused on the absence of dependencies of local-maximum confidence on the ideal occurrence/window widths in synthetic random datasets, and the other on the applicability of the episode-rule mining algorithms with a set of earthquake data.