Short-term prediction of intersection turning volume using seasonal ARIMA model

ABSTRACT

Accurate prediction of turning volumes at urban intersections is useful in Advanced Traffic Management Systems (ATMS). Existing studies on traffic flow prediction have mostly focused on midblock sections and only limited studies were undertaken on urban intersections. In the present study, models based on Seasonal Autoregressive Integrated Moving Average (SARIMA) were developed to predict the direction-wise turning volumes at an unsignalized three-leg intersection. Preceding three days of direction-wise turning volumes were used in the model to predict the following day’s turning volumes. Short term prediction of turning volumes was also experimented using both historic (preceding three days data) and real time data on the day of prediction. The results were promising with Mean Absolute Percentage Error (MAPE) of less than 10 in majority of the cases. The prediction scheme requires only limited data as input and open source software package R for estimation of model parameters and prediction.

KEYWORDS:

• Intersection
• turning volume
• prediction
• SARIMA
• Advanced Traffic Management Systems (ATMS)
• Intelligent Transportation Systems (ITS)

Disclosure statement

No potential conflict of interest was reported by the authors.