Non-linear model optimization for predicting time series is a challenge problem. In Intelligent Transportation Systems (ITS) application, the indispensable short-term traffic flow prediction with big data makes the problem worst. To improve the prediction accuracy and ensure real-time performance in the big data environment, we propose a novel co-evolutionary artificial bee colony (ABC) improved by differential evolution (DE) optimization algorithm combined with a traffic flow predicting model trained by extreme learning machine (ELM) neural network. The proposed model can inherit the better generalization performance and the less training time consumption of the standard ELM, and can achieve a more balanced search strategy with the optimized weights and biases to overcome the random initialization deficiency of the typical ELM, and successfully obtain higher prediction accuracy compared with state-of-the-art methods. To verify the efficiency of the proposed model, we apply it to Lozi and Tent chaotic time series simulations and measured traffic flow time series experiments. Simulation and experimental results demonstrate that the proposed model has superior performance and competitive computational efficiency.

用于预测时间序列的非线性模型优化是一个难题。在智能交通系统（ITS）应用中，不可缺少的大数据短期交通流量预测使问题变得最严重。为了提高预测精度并确保在大数据环境中的实时性能，我们提出了一种新的共进化人工蜂群（ABC），该算法通过差分进化（DE）优化算法与通过极端学习训练的交通流预测模型相结合来改进机器（ELM）神经网络。所提出的模型可以继承标准ELM的更好的泛化性能和更少的训练时间消耗，并且可以通过优化的权重和偏差实现更均衡的搜索策略，从而克服典型ELM的随机初始化缺陷，与最先进的方法相比，成功地获得了更高的预测精度。为了验证所提出模型的效率，我们将其应用于Lozi和Tent混沌时间序列模拟和实测交通流时间序列实验。仿真和实验结果表明，该模型具有优越的性能和具有竞争力的计算效率。