2013 Session: 530

• Development of Various Artificial Neural Network Car-Following Models with Converted Data Sets by Self-Organization Neural Network
  Abstract: This paper presents the development of a car-following model using the multilayer artificial neural network (ANN) structure. Four ANN car-following models were developed with various input variables in the car-following behavior. Tens of thousands of data points were used for the model developments, including acceleration from start, deceleration to stop, and mid to high speed car-following conditions on a test track. A four-layer neural network structure was set up and a genetic algorithm (GA) was utilized to determine the initial synaptic weights among the neurons based on the observed data sets. Back-propagation methodology was then utilized for fine tuning the synaptic weights further, however the models sometimes had a difficulty in learning such enormous number of raw data points. Therefore, a methodology of data point conversion was developed with, Kohonen Feature Map (KFM), a self-organization neural network model. The converted fewer data points were used for training the models. The results with and without data conversion were compared. In order to evaluate the ANN models, the existing well-known car-following model, the GM model, was calibrated with the same data sets. This paper concluded that the ANN models were successfully developed with KFM data conversion without deteriorating the original data quality. One of the four ANN models performed better than the GM car-following model. In comparing the results among the four ANN models, it was implied that the accelerations of the following vehicle and leading vehicle can also become key input variables for improving the modeling of car-following behavior.Keywords: Car-Following Model, Artificial Neural Networks (ANN), Back-propagation, Kohonen Feature Map (KFM), Genetic Algorithm (GA), Data Sampling, Microscopic Traffic Mode
  Authors: Tanaka, Mitsuru; Nakatsuji, Takashi
  Authors: Tanaka, Mitsuru; Nakatsuji, Takashi
  Year: 2013
  Document Type: Paper
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-0787
• Modeling of Spatial Transit Mode Choice Based on Smart Card Data in Seoul
  Abstract: In this study, using public transport smart card data observed in Seoul, Korea, we provide empirical evidence for the existence of a spatial correlation among transit mode choices at the Traffic Analysis Zone TAZ level. The Bayesian hierarchical modeling framework was applied to construct both spatial associations among TAZ levels and among commonly used utility functions with travel time and fare. We consider the binomial regression model with spatial effect by using a conditional autoregressive model (CAR) and regard a passenger¡¯s choice of metro rather than bus transport as a reference category. The results show that the areas with a higher probability that passengers will choose a bus are clustered and that those regions have fewer metro stations than bus stations. We also found the spatial correlation is statistically meaningful and potentially useful in the modeling of a spatial transit mode choice. Consequently, a reliable spatial interaction would constitute valuable information for transportation agencies in terms of their route planning and scheduling based on the transit smart card data.
  Authors: Eom, Jin Ki; Park, Man Sik; Heo, Tae-Young; Stone, John R.
  Authors: Eom, Jin Ki; Park, Man Sik; Heo, Tae-Young; Stone, John R.
  Year: 2013
  Document Type: Paper
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-1142
• Transit Smart Card Data Mining for Passenger Origin Information Extraction
  Abstract: The Automated Fare Collection (AFC) system, also known as the transit smart card system, has gained more and more popularity among transit agencies worldwide. Compared with the conventional manual fare collection system, an AFC system has its inherent advantages in low labor cost and high efficiency for fare collection and transaction data archival. Although it is possible to collect highly valuable data from transit SC transactions, substantial efforts and methodologies are needed for extracting such data because most AFC systems are not initially designed for data collection. This is especially true for Beijing’s AFC system, where a passenger’s boarding stop (origin) on a flat-rate bus is not recorded on the check-in scan. To extract passengers’ origin data from recorded SC transaction information, a Markov chain based Bayesian decision tree algorithm is developed in this study. Using the time invariance property of Markov chain, the algorithm is further optimized and simplified to reduce its computational complexity to linear. This algorithm is verified with transit vehicles equipped with GPS data loggers. Our verification results demonstrated that the proposed algorithm is effective and efficient in extracting transit passengers’ origin information from SC transactions with a relatively high accuracy. Such transit origin data are highly valuable for transit system planning and route optimization.
  Authors: Ma, Xiaolei; Wang, Yinhai; Chen, Feng; Liu, Jianfeng
  Authors: Ma, Xiaolei; Wang, Yinhai; Chen, Feng; Liu, Jianfeng
  Year: 2013
  Document Type: Paper
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-2156
• Handling Uncertainty in Transit Project Evaluation and Rating Process: A Comparison between the Existing FTA Approach and a Fuzzy Inference Approach
  Abstract: A fuzzy inference approach that ranks proposals for major transit projects is proposed, and its performance is compared with the existing approach used by the US Federal Transit Administration’s (FTA) New Starts Program. The FTA’s approach uses a rigid mathematical process in which 24 attributes of a proposal are scored initially. These scores are aggregated multiple times to obtain a single overall rating for the proposal, on which the proposal’s funding recommendation is made. In this approach, a small difference in the initial score can make significant differences in the final score; the final score is not stable with respect to small perturbations in the initial scores of the attributes. In any evaluation, there is always room for subjective judgment and associated uncertainty to enter, when determining the score of an attribute, when determining the breakpoints on the scoring scale, and when determining the value of the weight of the attributes. The proposed fuzzy inference approach incorporates fuzziness and approximation that is associated with the evaluation process and preserves it through the calculation process. The two approaches, the FTA’s and the proposed one, are compared using the data in the 2012 and 2013 reports of FTA’s New Starts Program proposal evaluation. The proposed fuzzy inference approach is found to be robust in dealing with evaluator’s uncertainty in the initial scores. The final score is found to be more stable than the FTA method, with respect to small changes in the initial values, weights, and breakpoint on the performance scale.
  Authors: Kikuchi, Shinya; Kronprasert, Nopadon
  Authors: Kikuchi, Shinya; Kronprasert, Nopadon
  Year: 2013
  Document Type: Paper
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-3173
• Urban Traffic State Estimation for Signal Control UsingMixed Data Sources and Extended Kalman Filter
  Abstract: This paper describes a methodology for fusing data from multiple sensors, including wireless de-vices and inductive loops, to make an estimation of the instantaneous state of an urban trafficnetwork. An extended Kalman filter is employed along with a state evolution model to make es-timates of the state in a discretized network. The instantaneous state is an estimate of the currentdistribution of vehicles in the network and their instantaneous speeds. Microsimulation tests wereused to evaluate the performance of the state estimation on a small urban networks. These resultsindicate low error between the estimated state and the known ground truth.
  Authors: Box, Simon
  Authors: Box, Simon
  Year: 2013
  Document Type: Paper
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-4310
• Modeling of Spatial Transit Mode Choice Based on Smart Card Data in Seoul
  Authors: Eom, Jin Ki
  Authors: Eom, Jin Ki
  Year: 2013
  Document Type: Presentation
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-1142
• Transit Smart Card Data Mining for Passenger Origin Information Extraction
  Authors: Ma, Xiaolei
  Authors: Ma, Xiaolei
  Year: 2013
  Document Type: Presentation
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-2156
• Development of Various Artificial Neural Network Car-Following Models with Converted Data Sets by Self-Organization Neural Network
  Authors: Tanaka, Mitsuru
  Authors: Tanaka, Mitsuru
  Year: 2013
  Document Type: Presentation
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-0787
• Handling Uncertainty in Transit Project Evaluation and Rating Process: A Comparison between the Existing FTA Approach and a Fuzzy Inference Approach
  Authors: Kronprasert, Nopadon
  Authors: Kronprasert, Nopadon
  Year: 2013
  Document Type: Presentation
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-3173
• Urban Traffic State Estimation for Signal Control Using Mixed Data Sources and Extended Kalman Filter
  Authors: Box, Simon
  Authors: Box, Simon
  Year: 2013
  Document Type: Presentation
  Subject: Data and Information Technology
  Session: 530
  Paper Number: 13-4310