Introducing Concept of Fuzzy Support Matrix for Interestingness Measures

Fuzzy association rules with its linguistic annotations and human interpretable form, has provided a convenient extension of association concepts to quantified attributes. The applicability is extended by combining extraction of both positive and negative association rules. Interestingness measures are used to filter out the useful and correct set of actionable association rules from the larger set of rules mined by association rule mining algorithms. Many measures such as Support, Confidence, Conviction and Certainty Factor, with their own area of applicability and statistical significance are popular. The wide range of measures is usually based on frequency counts or probability of occurrence of certain attribute patterns. Binary attributes uses a 2×2 contingency table as the basis for defining different measures. This paper presents concept of fuzzy support matrix using fuzzy partitions, as a natural extension of contingency table for the different interestingness measures. Those can be defined in a uniform and consistent manner. It uses the existing interestingness measures defined in new form using fuzzy support and illustrate these concepts using known data sets. This paper represent active research directions aimed at advancing the capabilities, applicability, and efficiency of fuzzy association rule mining in handling modern data challenges across various domains.