Rule discovery with a multi objective cultural algorithm

Abstract

Cultural algorithms (CA) are evolutionary systems which utilize agent technology and which supports any evolutionary strategy like evolutionary algorithm or swarm intelligence or ant algorithms. CA uses a basic set of five knowledge sources (KS’s) which are used in various animal species to guide the search toward best solutions and thus are better than evolutionary algorithms which are memory less blind search methods. The preserved knowledge in CA is disseminated throughout the system in future generations. Cultural algorithms have been used effectively in solving optimization problems, in engineering rule based systems, and combined with data mining to study complex social systems. However application of cultural algorithm for multi objective optimization of classification rules is hardly found in the literature. Research gap exists in using Cultural Algorithm for rule mining taking the various properties of rules as objectives for optimization. In the current study a cultural algorithm framework is proposed for rule mining considering it as a multi objective optimization problem.