Recently developed ant colony optimization metaheuristic procedure has been recast as a rule based machine learning method, called as ant colony classifier system, and applied to three process engineering examples. The learning algorithm addresses the problem of knowledge acquisition in terms of rules from example cases by developing and maintaining the knowledge base through the use of simple mechanism, pheromone trail information matrix and use of available heuristic information. The performance of an ant colony classifier is compared with the well-known decision tree based C4.5 algorithm in terms of the predictive accuracy on test cases and the simplicity of rules discovered. The results indicate that the ant classifier is able to discover rules in the data sets with better predictive accuracy than the C4.5 algorithm. © 2004 Elsevier Ltd. All rights reserved.