FCACO: Fuzzy classification rules mining algorithm with ant colony optimization

Ant colony optimization (ACO) is relatively new computational intelligence paradigm and provides an effective mechanism for conducting a global search. This work proposes a novel classification rule mining algorithm integrating ACO for search strategy and fuzzy set for representation of the rule terms to give the system flexibility to cope with continuous values and uncertainties typically found in real-world applications and improve the comprehensibility of the rules. The algorithm uses a strategy that is different from 'divide-and-conquer' and 'separate-and-conquer' approaches used by decision trees and lists respectively; and simulates the ants' searching different food sources by using attribute-instance weighting and an effective pheromone update strategy for mining accurate and comprehensible rules. Obtained results from several real-world data sets are analyzed with respect to both predictive accuracy and simplicity and compared with C4.5Rules algorithm.