Journal of Digital Information Management L. Yang, K.S. Li, W.S. Zhang, Y. Wang, Z.X. Ke 13 5 2015 http://dline.info/fpaper/jdim/v13i5/v13i5_1.pdf A new classification algorithm based on multi-ant and pheromone repulsion principles is studied and proposed in this paper to improve the prediction accuracy of the classification rule based on the traditional Ant-Miner algorithm. The proposed algorithm uses multiant colony construction method to reduce dependence on random initial term. The volatile coefficient of pheromone update is added to prevent premature convergence. The new state transition function based on pheromone repulsion principle is also presented to avoid falling into local optimum, whereas a rule quality formula is improved. Several aspects of the proposed algorithm are investigated by experimenting of benchmark data sets from the University of California at Irvine repository. We study the performance of our proposed approach and compare it with several commonly used classification algorithms, such as Ant-Miner and Ant-Miner with graphical user interface. Experimental results indicate that the proposed approach builds more accurate models than the compared algorithms. The high accuracy supplemented by the comprehensibility of the discovered rule sets is the main advantage of this method.A new classification algorithm based on multi-ant and pheromone repulsion principles is studied and proposed in this paper to improve the prediction accuracy of the classification rule based on the traditional Ant-Miner algorithm. The proposed algorithm uses multiant colony construction method to reduce dependence on random initial term. The volatile coefficient of pheromone update is added to prevent premature convergence. The new state transition function based on pheromone repulsion principle is also presented to avoid falling into local optimum, whereas a rule quality formula is improved. Several aspects of the proposed algorithm are investigated by experimenting of benchmark data sets from the University of California at Irvine repository. We study the performance of our proposed approach and compare it with several commonly used classification algorithms, such as Ant-Miner and Ant-Miner with graphical user interface. Experimental results indicate that the proposed approach builds more accurate models than the compared algorithms. The high accuracy supplemented by the comprehensibility of the discovered rule sets is the main advantage of this method.A new classification algorithm based on multi-ant and pheromone repulsion principles is studied and proposed in this paper to improve the prediction accuracy of the classification rule based on the traditional Ant-Miner algorithm. The proposed algorithm uses multiant colony construction method to reduce dependence on random initial term. The volatile coefficient of pheromone update is added to prevent premature convergence. The new state transition function based on pheromone repulsion principle is also presented to avoid falling into local optimum, whereas a rule quality formula is improved. Several aspects of the proposed algorithm are investigated by experimenting of benchmark data sets from the University of California at Irvine repository. We study the performance of our proposed approach and compare it with several commonly used classification algorithms, such as Ant-Miner and Ant-Miner with graphical user interface. Experimental results indicate that the proposed approach builds more accurate models than the compared algorithms. The high accuracy supplemented by the comprehensibility of the discovered rule sets is the main advantage of this method.