一种基于双周期时间序列的短时交通流预测算法

A Short-term Traffic Flow Forecast Algorithm Based on Double Seasonal Time Series

根据城市道路短时交通流特征,在ARIMA(autoregressive integrated moving average model)模型和SARIMA(seasonal autoregressive integrated moving average model)模型的基础上,提出一种既满足城市道路日周期性和周周期性的短时交通流预测模型DSARIMA(double seasonal autoregressive integrated moving average model)模型,并根据城市道路工作日与非工作日交通流特点,提出该模型的预测算法。该算法采用两种方式利用ARIMA模型进行交通流预测,一种方式采用当前时刻前N1段时间进行预测,另一种方式采用当前时刻前N2天同一时段的交通流进行预测,并用改进的贝叶斯模型算法根据两种预测结果与实际值的误差来确定该种方式的权值,最后的预测结果为两种方式预测结果与其权值乘积之和。实验结果表明,该模型在交通流预测上,相比ARIMA模型和SARIMA模型预测具有更好的平稳性与更高的预测精度。

According to the characteristics of urban road short-term traffic flow,one kind of short-term traffic flow forecast model named DSARIMA（ double seasonal autoregressive integrated moving average model） was proposed,which can meet the requirements of predicting daily pattern and weekly pattern traffic flow in urban road based on the ARIMA（ autoregressive integrated moving average model）model and SARIMA（ seasonal autoregressive integrated moving average model） model. According to the characteristics of the traffic flow in weekdays and weekends in urban road,a forecast algorithm was given which use two methods to forecast traffic flow by ARIMA model,one forecast using N1time quantum before the given time,the other forecast using the same time quantum of N2days before the given time,and use improved BAYESIAN model to decide the powers of the two methods by calculating the difference of the forecast results and the real results. The final forecast results were the sum of the respective results of the two methods multiplied by its power. The results showed that the DSARIMA model has better stability and are more accurate than the ARIMA model or SARIMA model when applied to the short-term traffic flow forecast.