Abstract: Aimed at traveling salesman problems， a novel discrete artificial bee colony algorithm is proposed. Based on the characteristics of such problems and discrete variables， the transforming mechanism and searching strategy of leader bees， follower bees and scout bees are redefined. The roles of bees are changed dynamically according to the values of profitability ratios. The 2-Opt operator and learning operator are used for leader bees to search the neighborhood of food sources so as to accelerate the convergence. A taboo list is introduced for follower bees to improve the algorithm’s intensification ability， and a repulsion operator is designed for scout bees to maintain the diversity of bee colonies. The proposed algorithm can strike a good balance between exploration and exploitation by using these operators. The simulation results show that it can improve the efficiency of solving traveling salesman problems by finding relatively satisfactory solutions in a short time.

Key words: discrete artificial bee colony algorithm, TSP(traveling salesman problem), 2-Opt, learning operator, repulsion operator