Volume 18, Issue 4 (2018)                   MCEJ 2018, 18(4): 115-129 | Back to browse issues page

XML Persian Abstract Print


Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

Seyedabrishami S, Mohades Deylami A, amini V, Iranmanesh M. Short-term prediction of passenger demand in bus stations, case study: Karimkhan bridge- Jomhoori square. MCEJ. 2018; 18 (4) :115-129
URL: http://mcej.modares.ac.ir/article-16-15425-en.html
1- Assistant Professor of Transportation Engineering
2- PhD Candidate in computer Engineering, Electrical and Computer Engineering Dept.,Tarbiat Modares University
3- Master graduated in Transportation Planning, Tarbiat Modares University, Tehran, Iran
4- Master Graduated in Transportation Planning,Tarbiat Modares University
Abstract:   (3425 Views)
Nowadays, cities as a place of living and human activity are facing serious challenges in providing human needs. Increasing in population growth, vehicle ownership and communication development has led to complexity of the transportation system and its problems, including congestion, environmental pollution and the consumption of non-renewable resources. Therefore, changes in urban transport policies and efforts to develop and more use of the public transport, especially the bus, are one of the most important concerns in urban transport planning. A review of various studies suggests that planning for efficient use of bus infrastructures and enhancing the efficiency of public transportation operation in the world, require information on the infrastructure and passenger demand for lines and bus stations. Accordingly, it is necessary to carry out studies to predict passenger demand for bus stations in Tehran. Thus, this study predicts bus stations passenger demand for future short-term periods, using data gathered by AFC (Automated Fare Collection) and AVL (Automatic Vehicle Location). For this purpose, firstly AFC and AVL data was sorted according to the time for each bus line. Since passengers use their smart card while they are getting off the bus it means at the exit station thus identifying their origin station is vital, so that in second step, data of two data bases is compared and matched by writing computer code in Matlab software to determine the origin stations of passengers and then forming origin-destination demand matrix for each bus line in terms of its stations. This matrix is considered as the main data base of the study, a time series analysis, a seasonal autoregressive integrated moving average (SARIMA) and neural network as an artificial model are calibrated based on the available data. Both models’ goodness of fit indices are compared in terms of learning and generalization capabilities. For this purpose, initial data is divided into two subsets called learning and test data sets and comparison indices are computed for both aforementioned sets. The models’ results show that the multi-layer perceptron neural network model in terms of goodness of fit indices in both learning and generalization capabilities in prediction of bus station passenger demand is better than SARIMA model; however, the manner of influencing different factors such as day of week or month of year in passenger demand in each station is more clear in time series analysis. The passenger demand for each stations in first month in spring is different from the rest months in this season. Months in summer is also show different trends for passenger demand, while all months in fall and the first two months in winter have similar passenger demand in various stations. Official holidays has also significant influence on passenger demand so that reduce passenger demand by approximately 256 persons on average. All days in week have meaningful effects on passenger demand in comparison with Friday so that Monday and Thursday have the highest and the lowest effect on weekday passenger demand in bus stations in comparison with Friday, respectively. This analysis comparison show that if the precision of future prediction is important then neural network outweigh time series regression, while the identification of influential variables on passenger demand is better done by time series analysis.
Full-Text [PDF 795 kb]   (1828 Downloads)    
Article Type: Original Manuscript | Subject: Earthquake
Received: 2017/08/11 | Accepted: 2018/11/11 | Published: 2018/11/15

Add your comments about this article : Your username or Email:
CAPTCHA

Rights and permissions
Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.