Abstract: Short-term passenger flow forecasting is considered as a vital component of transportation systems which can aid in fine-tuning travel behaviors, reducing passenger congestion, supporting transportation system management and in enhancing service quality of transportation systems. For the same reason, passenger flow prediction still remains as an active research area. Many researchers have developed various travel demand prediction models based on different methodologies. Since the passenger flow prediction is complicated and non-linear in nature, the non-parametric technologies such as Neural Network, Support Vector Machine, Kalman filtering, Random Forest have been applied in passenger flow prediction. The deep-structured architecture of deep neural networks (DNN) could extract complex structure and build internal representation from the inputs, which is expected to outperform the traditional travel demand prediction models for passenger flow. This paper presents a deep neural network (DNN) based prediction model for predicting hourly passenger flow in an urban metro rail system. The input features of the model include temporal features like the day of a week, hour of a day and the holidays and spatial features include the respective metro stations and the passenger flow. These features are combined and multiple scenarios are modelled. The models are applied and evaluated with the passenger flow data from the metro system. Mean absolute error (MAE) and root mean squared error (RMSE) are used as measures of performance of these models. The experimental results showed that the DNN based prediction models effectively captures the non-linear relationship between the influential input features and the passenger flow. The model has the capability to provide an accurate and universal metro rail passenger flow prediction.
Keywords: Short-term passenger flow, Deep Neural Network (DNN), Metro Rail System, Root Mean Squared Error (RMSE).
Title: Short Term Passenger Flow Prediction Using Deep Neural Network: A Case Study of Urban Metro Rail System
Author: Girish H., Dr. K. Sunil Kumar
International Journal of Interdisciplinary Research and Innovations
ISSN 2348-1218 (print), ISSN 2348-1226 (online)
Research Publish Journals
