Abstract: Aiming at the disadvantages that Artificial Bee Colony algorithm（ABC） easily traps in the local optimum and has slow convergence rate, the present-optimal food source and the function of the inertia weight are introduced to improve the update forms of the food source. The parameter selection of SVR transforms to a combinatorial optimization problem, uses the improved artificial bee colony algorithm to optimize it and therefore the prediction model with SVR optimized by improved ABC is obtained. Taking the short-term traffic flow data as an example, the predictive results of the model are compared with ACO-SVR, PSO-SVR and ABC-SVR, the results show that the model is superior to the predictive effect of the other three, its run time is the shortest and it has better learning ability and generalization ability.

Key words: artificial bee colony algorithm, support vector regression, traffic-flow prediction, ant colony algorithm, particle swarm algorithm

摘要： 针对人工蜂群算法存在的易陷入局部最优、收敛速度慢的缺点，引入当前最优食物源和惯性权重函数，对该算法的食物源更新方式进行改进;针对支持向量回归机的参数优化问题，将其转化为组合优化问题，并使用改进的人工蜂群算法进行优化求解，进而得到人工蜂群算法优化SVR的预测模型。以短期交通流量数据为例，将该模型的预测结果与蚁群算法优化的支持向量回归机（ACO-SVR）、粒子群算法优化的支持向量回归机（PSO-SVR）和未改进的蜂群算法优化的支持向量回归机（ABC-SVR）进行对比分析，结果表明该模型的预测效果最优且运行时间最短，具有更好的学习能力和推广能力。

关键词: 人工蜂群算法, 支持向量回归机, 交通流量预测, 蚁群算法, 粒子群算法