基于关联维的事故时间序列分形特征分析

根据我国"十五"期间事故统计数据,构造了一个事故时间序列;基于相空间重构理论,对事故序列进行了相空间重构,应用G-P算法,对事故时间序列的分形特征进行了研究,计算得出事故时间序列的关联维数。研究结果表明:当嵌入维数达到8以后,该事故动力学系统具有稳定的关联维数4.7,说明至少有5个因子在影响着事故时间序列的动态变化,并且该系统的有效自由度为8。本研究对建立事故时间序列的预测模型有较大的参考价值。

Analysis on Fractal Features of Accident Time Series Based on Correlation Dimension

Accident time series were constructed according to the accidents data in China during the "Tenth Five-Year" period.Based on the phase-space reconstruction theory,phase-space of accident time series was reconstructed.The fractal features of accident time series were studied by using G-P(Grassberger-Procaccia)arithmetic and correlation dimension of accident time series was calculated out.The results indicate that the dynamic system of accidents has a steady correlation dimension of 4.7 after the embedded dimension comes to 8,which explains that at least five factors are controlling the dynamic changes of accidents time series,and the effective freedom is 8.The research provides much reference value for establishing predictable models of accident time series.