论文部分内容阅读
道路运行车速预测是交通预测的难点,运行车速随交通条件的变化而变化,为提高道路运行车速预测精度,构建了自回归移动平均(ARIMA)时间序列模型,并结合实例对重庆市江北区红黄路早高峰小客车平均运行车速进行了预测。结果表明:相较于传统的线性回归、多项式拟合、指数拟合和模糊线性回归预测模型,ARIMA预测模型的平均绝对误差分别下降了19.1%,50%,6.5%和3.7%;另外,将原始序列取自然对数后再建立ARIMA的Log-ARIMA模型可进一步提高预测精度,预测绝对误差为5.21,与普通ARIMA模型比较,平均相对误差下降了29.9%。
Road speed forecasting is a difficult point in traffic forecasting. The running speed varies with traffic conditions. In order to improve the prediction precision of road running speed, an autoregressive moving average (ARIMA) time series model is constructed. The average running speed of the early peak passenger cars in the Yellow Road was predicted. The results show that the average absolute error of ARIMA prediction model decreases by 19.1%, 50%, 6.5% and 3.7% respectively compared with the traditional linear regression, polynomial fitting, exponential fitting and fuzzy linear regression prediction model. In addition, After the natural logarithm of the original sequence was taken, Log-ARIMA model of ARIMA was established to further improve the prediction accuracy. The predicted absolute error was 5.21. Compared with the common ARIMA model, the average relative error decreased by 29.9%.