基于ARIMA模型的短时交通流量预测算法研究

Study of short-term traffic flow forecast algorithms based on ARIMA model

针对造成基于线性最小方差预报原理的Astrom算法在多步预测过程中误差逐步增大的原因,通过增加误差动态修正因子,提出一种改进的短时交通流量预测算法.该算法基于ARIMA模型结构的时间序列分析方法,采用矩估计法进行参数初估计,用最小二乘法进行参数精估计,用BIC准则为模型定阶.对大量实测数据进行仿真实验,对多个统计量进行误差分析.结果表明,改进算法在应用于时变性强的短时交通流量预测时,相对于Astrom算法具有更好的预测性能.

Point to the factors making Astrom algorithm based on the linear smallest variance forecast principle that the probable error increased gradually in many step forecast process, and the error dynamic modification factor in this algorithm foundation was increased, then one kind of improved short-time traffic flow forecast algorithm was proposed. This algorithm was based on time series analysis method adopting ARIMA model structure, moment estimation method was used for parameter wide estimation, least square method was adopted for parameter exactly estimation, and the BIC rule was used for determining the order of the model number. A lot of real observation data were used for simulation tests and a number of statistics were introduced to analyze the errors. The results show that the improved algorithm has better forecast performance than Astrom algorithm when applied to the short-term variability strong traffic.