SHrimp: Descriptive Patterns in a Tree
Sibylle Hess, Nico Piatkowski, Katharina J. Morik · 2014
Abstract. The appliance of the minimum description length (MDL) principle to the field of theory mining enables a precise description of main characteristics of a dataset in comparison to the numerous and hardly understandable output of the popular frequent pattern mining algorithms. The loss function that determines the quality of a pattern selection with respect to the MDL principle is however difficult to analyze and the selection is computed heuristically for all known algorithms. With SHrimp, the attempt to create a data structure that reflects the influences of the pattern selection to the database and that enables a faster computation of the quality of the selection is initiated.