Abstract
This paper aims to develop rolling grey forecasting models (RGM) to predict short-term traffics. Two types of RGM models are developed and compared: RGM(1,1) and RGM(1,N). To investigate and validate the accuracy and applicability of proposed models, two time horizons of short-term traffics of 1-minute and 5-minute are applied, respectively. For comparison, two commonly used short-term traffic prediction models: statistical timeseries model (ARIMA) and artificial neural network (ANN), are also developed. The accuracies in term of mean absolute percentage error (MAPE) of various rolling intervals (4-8 intervals) and prediction periods (1-5 periods) of the proposed model are also compared. The results show that both of RGM(1,1) and RGM(1,N) perform better at fewer rolling interval and prediction period. Besides, RGM(1,6) remarkably outperforms in predicting three traffics, followed by RGM(1,1). Obviously, the performances and applicability of proposed RGM models are validated.
