论文部分内容阅读
对随机性较强的公路交通事故,提出了采用灰色残差模型来减小预测误差,模型用多个指数函数的线性叠加来分段描述事故预测值,克服了灰色模型只能描述单调变化的过程,仅适用于预测具有较强指数变化规律事故序列的缺点。对影响灰色残差建模的修正对象、数据选取、数据预处理和修正方法等多种因素作了详细分析和划分,并归纳出4种实用的灰色残差模型。实例分析结果表明:与灰色模型相比,各种灰色残差模型的预测误差降低了70%~80%,其中采用残差直接建模、一次还原的灰色残差模型,复杂程度低,预测误差小于5%。
For the road traffic accident with strong randomness, this paper proposes to use the gray residual model to reduce the prediction error. The model uses the linear superposition of multiple exponential functions to describe the accident prediction value segment by section, overcomes that the gray model can only describe the monotonic change The process is only applicable to the shortcomings of predicting sequences of accidents with strong exponential behavior. The paper analyzes and divides a variety of factors, such as the object of amendment, data selection, data preprocessing and correction methods, which affect the modeling of gray residuals, and draws four practical models of gray residuals. The results of the example analysis show that compared with the gray model, the prediction error of various gray residual models is reduced by 70% ~ 80%. The gray residual model with direct modeling of residuals, primary reduction, low complexity and prediction error Less than 5%.