An improved multiple ant colony algorithm was presented which aims at load balance and optimization of network resources management. Through the interaction and dynamic update among the pheromone of multiple ant colony which are on behalf of the network traffic, the algorithm enabled network traffic to share a number of paths available; Through the combination of orientable choice and stochastic choice to select best path self-adaptively, the algorithm achieved the traffic load balancing; Through setting the maximum-minimum pheromone to avoid premature convergence, the algorithm increased capabilities of global optimum search; Through the improvement of the cost function and the comprehensive use of a variety of improved ways, the algorithm improved self-adaption of the algorithm. The results of simulation experiment demonstrate that compared with multiple ant colony load balance algorithm，improved multiple ant colony algorithm has superiority in reducing time of auto adaption, lowering packet loss rate and improving efficiency of load balance.