Development of Short-term Traffic Volume Prediction Models for Adaptive Traffic Control

Abstract

This paper aims at establishing an accurate short-term traffic volume prediction method by introducing calculations and characteristics of triple moving average method, single exponential smoothing method, and double exponential smoothing method. Based on field data collected from intersections in Fuzhou, the prediction accuracy of three prediction methods above was calculated respectively at 5-minute, 10-minute and 15-minute statistics intervals and analyzed. Analysis results indicate that prediction methods and statistics intervals have significant impacts on the prediction accuracy. To be specific, the prediction accuracy of double exponential smoothing method is higher than other two methods; The accuracy at 10-minute statistics intervals is higher than that at other two intervals. The research results will provide an accurate volume prediction method for adaptive traffic control.