基于近似熵的交通流序列趋势变化检测

交通流趋势变化特征分析是交通流预测的基础．为了提取交通流序列随时间推移所呈现出来的宏观变化规律，提出了一种用于检测交通流序列趋势变化的滑动移除近似熵方法．通过对交通流序列趋势规律进行研究，首先将其细分为上升趋势、平稳波动趋势、下降趋势，然后根据不同趋势变化的时间序列复杂程度不同，建立了滑动移除近似熵方法求解其滑动移除近似熵的值，并根据得到的时间序列提取交通流序列趋势变化．最后以北京市四环路某一断面交通流序列为例，用建立的模型对交通流序列趋势变化进行检测，并与滑动t检验方法结果对比．研究结果表明本文提出的方法能够对交通流序列趋势变化进行检测，且检测结果与实际交通流序列趋势变化比较吻合，研究结论可为短时交通流预测建模提供拳者依据．

Trend change detection of traffic flow based on approximate entropy

The analysis of traffic flow characters is the basic of traffic flow to forecast. In order to ex- tract trend change of traffic flow, a detection approach is proposed to detect the trend change of traffic flow. The research divides the trend of short-term traffic flow into three phases： the trend of down- ward phase, the stable fluctuation phase, and the trend of rising phase. Then the research points out that trend change of traffic flow reflects different dynamic characteristics, and establishes moving cut data approximate entropy （MC-ApEn） to detect the trend change of traffic flow. Finally, the empiri- cal researches proceed by using traffic parameter data from the road network, which is compared to the moving t test method. The results show the proposed methodology can detect trend change of traffic flow and the accuracy is improved, and the results could provide supports for traffic flow forecasting.