Real-time short-term forecasting method of remaining parking space in urban parking guidance systems

Abstract

Short-term forecasting of the remaining parking space is important for urban parking guidance systems (PGS). The previous methods like polynomial equations and neural network methods are difficult to be applied in practice because of low accuracy or lengthy initial training time which is unfavourable if real-time training is carried out on adapting to changing traffic conditions. To forecast the remaining parking space in real-time with higher accuracy and improve the performances of PGS, this study develops an online forecasting model based on a time series method. By analysing the characteristics of data collected in Nanjing, China, an autoregressive integrated moving average (ARIMA) model has been established and a real-time forecasting procedure developed. The performance of this proposed model has been further analysed and compared with the performances of a neural network method and the Markov chain method. The results indicate that the mean error of the proposed model is about 2 vehicles per 15 minutes, which can meet the requirements for general PGS. Furthermore, this method outperforms the neural network model and the Markov chain method both in individual and collective error analysis. In summary, the proposed online forecasting method appears to be promising for forecasting the remaining parking space in supporting the PGS.