A prediction model of bus arrival time at stops with multi-routes

Accurate bus arrival time is fundamental for efficient bus operation and dispatching decisions. This paper proposed a new prediction model based on support vector machine (SVM) and artificial neural network (ANN) to predict bus arrival time at an objective stop with multi-routes. The preceding bus arrival time of objective route and all other routes passing the same stop, and travel speed of the target one are three inputs of the model. A case study was conducted with data collected in all workdays in October, 2014 in Zigong, Sichuan, China. The results of the proposed model indicate that both SVM and ANN models have high accuracy, while the ANN model is better than SVM model comparatively. The mean absolute percentage errors (MAPE) of prediction are less than 10% in most cases. By contrast, two groups with inputs changed or removed are set up to demonstrate the suitability of three inputs. No matter which method to use, SVM or ANN, the consequence of the proposed model is better than comparative groups. (C) 2017 The Authors. Published by Elsevier B.V.