Classification Rule Mining with an Improved Ant Colony Algorithm

摘要

This paper presents an improvement ant colony optimization algorithm for mining classification rule called ACO-Miner. The goal of ACO-Miner is to effectively provide intelligible classification rules which have higher predictive accuracy and simpler rule list based on Ant-Miner. Experiments on data sets from UCI data set repository were made to compare the performance of ACO-Miner with Ant-Miner. The results show that ACO-Miner performs better than Ant-Miner with respect to predictive accuracy and rule list mined simplicity.