Study on the Time Series of Traffic Flow Based on Algorithm Complexity Measure and Approximate Entropy

摘要

In order to quantitatively analyze the complexity of traffic flow system, algorithm complexity and approximate entropy are introduced, algorithm complexity of speed time series is used to estimate the ratio of the system’s cyclical component, and a plurality of divided intervalsare selected to improve the estimation ability of algorithm complexity in the reconstruction of sequence. In the calculation of approximate entropy, speed variation sequence obtained from speed sequence is used to remove the trend, and then approximate entropy of speed variation sequence isapplied to estimate system’s complexity in the changes of structure. Calculation of the actual data sequences of traffic flow shows that algorithmic complexity can be obtained when the sequence length is over 800, and approximate entropy can be obtained when the sequence length is over500; algorithm complexity and approximate entropy of traffic flow are low at a synchronizing state, increase at congested status and are the maximum at a free state. Therefore, different algorithm complexity and approximate entropy correspond to the traffic flow under different conditions,algorithm complexity can be used to analyze relatively longer traffic flow sequence, and approximate entropy can be applied to analyze relatively shorter traffic flow sequence.