An improved evolutionary approach for association rules mining

Abstract

This paper tackles the association rules mining problem with the evolutionary approach. All previous bio-inspired based association rules mining approaches generate non admissible rules. In this paper, we propose an efficient strategy called delete and decomposition strategy permits to avoid non admissible rules. If an item is appeared in the antecedent and the consequent parts of a given rule, this rule is composed on two admissible rules. Then, we delete such item to the antecedent part of the first rule and we delete the same item to the consequent part of the second rule. We also incorporate the suggested strategy into two evolutionary algorithms (genetic and mimetic algorithms), To demonstrate the suggested approach, several experiments have been carried out using both synthetic and reals instances. The results reveal that it has a compromise between the execution time and the quality of output rules. Indeed, the improved genetic algorithm is faster than the improved mimetic algorithm whereas the last one outperforms the genetic algorithm in terms of rules quality.