Bus Arrival Time Prediction Based on the k-Nearest Neighbor Method

In this paper, a modified k-nearest neighbor (k-NN) method integrating the cluster analysis and principal component analysis is applied to bus arrival time (BAT) prediction using historical bus GPS data. The methodology of the k-NN method based on processed GPS data is presented. To validate the introduced k-NN approach, empirical analysis based on data collected from the bus No. 16 of Beijing public transport holdings, Ltd. (BPT) is performed. The results show that the k-NN method has lower prediction error than the ANN model and is more powerful in the BAT prediction.