双曲正切函数在事故黑点鉴别中的应用

Application of hyperbolic tangent function for accident prone location identification

摘要:

为改善双指数函数y=a×ebx+c×edx在累计频率曲线拟合和事故黑点鉴别时存在的问题,提出用双曲正切函数y=a+b×tanh(cx+cd))作为替代,以提高黑点鉴别效果.通过仿真和实际数据验证对2种拟合算法进行对比分析.结果表明:累计频率曲线采用双曲正切函数进行拟合,其拟合效果好,事故黑点鉴别检出率较高,黑点鉴别的结果比较稳定,对不同事故统计年限的事故黑点鉴别有很好的适应性.此外,一定事故率水平下由累计交通量决定的事故次数对事故黑点的误检率有非常重要的影响.双曲正切函数可替代双指数函数进行事故黑点鉴别,同时为将误检率控制在20%左右,建议待鉴别道路的平均每公里事故次数不小于4次.

Abstract:

In order to improve the ability of the accident prone location identification,the hyperbolic tangent y=a+b×tanh(cx+cd) is proposed as a substitute fitting function of the double exponential function y=a×ebx+c×edx for fitting scatter diagram of accident cumulative frequency.The comparative analysis of the two fitting functions was conducted by simulation analysis and actual data confirmation.The results indicated that the hyperbolic tangent function has higher correlation coefficient in curve fitting,higher detection rate in accident prone location identification,more stable results and better adaptability for the cumulative traffic volume.More significant finding is proposed that the accident number,which decided by cumulative traffic volume in certain accident rate,has very important influence on misdetection rate of accident prone location identification.The hyperbolic tangent can be used for fitting scatter diagram of accident cumulative frequency instead of the double exponential function,and the suggested number of average accident per kilometer should not less than 4 to control the misdetection ratio to about 20%.