Abstract: Traffic flow prediction is a very important research area of intelligent transportation systems. Traditional prediction methods does not work very well because of the complexity of the influencing factors. So an ACS algorithm based on [t] distribution adaptive mutation is proposed. By setting dynamic variation control scales and multiple degrees of freedom to construct adaptive mutation algorithm. Overall optimization results can be obtained, which is better than Gaussian mutation and Cauchy mutation. And then short-time traffic flow forecasting method is proposed which based on BP neural network optimized by ACS. By optimizing the initial weights and threshold parameters, it can improve the short-term traffic flow prediction accuracy. The simulation results show that it can get better prediction results.

Key words: short-term traffic flow prediction, neural network, variability factor, improved cuckoo search algorithm

摘要： 交通流量预测是智能交通系统中非常重要的研究领域，因为交通流量的复杂性，传统的预测方法不能很好地预测。提出一种基于[t]分布自适应变异优化的布谷鸟算法，通过动态变异控制尺度和设置多个自由度来构造自适应变异算法，可以获得优于高斯变异和柯西变异的整体优化效果。在此基础上，提出改进布谷鸟搜索算法优化神经网络的交通流量预测模型（ACS-BPNN），通过优化BP神经网络的初始权值和阈值参数，以提高短时交通流量预测精度。仿真结果表明，该方法取得比较好的预测结果。

关键词: 交通流量预测, 神经网络, 变异尺度, 改进布谷鸟搜索算法