OCEC: a novel classification algorithm based on evolutionary computation

A novel classification algorithm, OCEC (Organizational CoEvolutionary algorithm for Classification), based on evolutionary computation for data mining is proposed. It is compared to GA-based and non GA-based algorithms on 8 datasets from the UCI machine learning repository. Results show OCEC can achieve higher prediction accuracy, smaller number of rules and more stable performance.

1 Introduction

Classification is one important task of data mining. Presently> the classification methods widely used are rough set, neural network, decision tree, genetic algorithm and so on. In GA-based classification methods, two main traditional ones are Michigan and Pittsburgh approach. Based on these two approaches, some hybrid approaches have been proposed[1] [2], and more preferable results are obtained.