Proposing an efficient combination of interesting measures for mining association rules via NSGA-II

Abstract

Selecting accurate and simple association rules that efficiently cover all data samples is very important in knowledge discovery. There are several measures to assess accuracy and relations in a rule. This poses a challenge for researchers to select effective measures. Combining different measures via multi-objective evolutionary algorithms is an effective method to select suitable association rules. Therefore in this paper NSGA-II algorithm is employed for rule selection via different combination of existing measures (support, certainty factor, change of support, Yao and Liu's one way support, cosine and lift) as objectives. The contributions of the paper are twofold. Firstly, some existing measures are modified. Secondly, several experiments are done to evaluate the performance of different combinations of measures through NSGA-II. The experimental results show that the combination of certainty factor and square of cosine measures are more effective in rule selection.

A Genetic Algorithm based framework for mining quantitative association rules without specifying minimum support and minimum confidence

“…With the approximation of cosine 2 value to zero, the majority of transactions possess item X, but lack item Y , and vice versa [50]. The square of this measure is used in this paper since it is more accurate [51]. It is de ned through Eq.…”