ARIMA models for bus travel time prediction

Abstract

In this paper, the time series model, Autoregressive Integrated Moving Average (ARIMA) is used to predict bus travel time. ARIMA model is simpler used for predicting bus travel time based on travel time series data (historic data) compared to regression method as the factors affecting bus travel time are not available in detail such as delay at link, bus stop, intersection, etc. Bus travel time prediction is an important aspect to bus operator in providing timetable for bus operation management and user information. The study aims at finding appropriate time series model for predicting bus travel time by evaluating the minimum of mean absolute relative error (MARE) and mean absolute percentage prediction error (MAPPE). In this case, data set was collected from the bus service operated on a divided 4-lane 2-way highway in Ipoh-Lumut corridor, Perak, Malaysia. The estimated parameters, appropriate model, and measures of model performance evaluation are presented. The analysis of both Ipoh to Lumut and Lumut to Ipoh directions is separately performed. The results show that the predicted travel times by using the moving average, MA(2) and MA(1) model, clearly fit with the observed values for both directions, respectively. These appropriate models are indicated by the minimum MARE and MAPPE values among the tentative models. It is concluded that MA(2) and MA(1) models are able to be appropriately applied in this case, and those models can be used for bus travel time prediction which helping in the timetable design or setup.