Title: Interestingness measures for quantified and ordered categorical attributes using fuzzy approach

Authors: Swati R. Ramdasi; Shailaja C. Shirwaikar; Vilas Kharat

Addresses: Department of Computer Science, Savitribai Phule Pune University, Ganeshkhind Pune 411 007, Maharasthra, India ' Department of Computer Science, Savitribai Phule Pune University, Ganeshkhind Pune 411 007, Maharasthra, India ' Department of Computer Science, Savitribai Phule Pune University, Ganeshkhind Pune 411 007, Maharasthra, India

Abstract: Fuzzy association rules with its linguistic annotations, provides a convenient extension of association concepts to quantified attributes and broadens their applicability by also extracting negative association rules. Traditionally, interestingness measures are used to filter out the right set of actionable association rules from the larger set of rules mined by algorithms. Several measures from support and confidence to conviction and certainty factor have their own area of applicability and statistical significance. This paper presents support matrix using fuzzy partitions, as a natural extension of contingency table for quantified and ordered categorical attributes so that interestingness measures can be defined in a uniform and consistent manner. The existing interestingness measures defined in new form are used to characterise complimentary and substitute attributes and new interestingness measures are proposed to measure the irrelevance of attribute data. Theoretical evaluation of various properties for interestingness measures helps in identifying representative set of eight measures.

Keywords: interestingness measures; association rules mining; fuzzy sets.

DOI: 10.1504/IJFCM.2019.100348

International Journal of Fuzzy Computation and Modelling, 2019 Vol.2 No.4, pp.353 - 381

Received: 30 May 2018
Accepted: 14 Jan 2019
Published online: 26 Jun 2019 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article