Past Seminars


Date:11 November 2003
Topic:A Brief Introduction to Data Mining with CART Decision Trees
Speaker:Dan Steinberg, Salford Systems
Abstract:Data mining is a fast developing research area in which a variety of highly effective modeling and data exploration tools have emerged. Although some major roots of data mining methodology can be traced back to applied econometricians in the 1960's (Morgan and Sondquist) and statisticians in the 1970's and early 1980's (Jerome Friedman, 1977; Breiman, Friedman, Olshen, and Stone, 1984), data mining has largely been the province of computer scientists and a separate breed of researchers blending machine learning, robotics, database, and statistical themes. This introduction is intended to give novices with classical statistical and econometrics backgrounds some insight into one of the most popular and accessible data mining method.
Date:11 November 2003
Topic:Performance & Analysis of Carriers Technologies & Strategies in Sequential Auction Transportation Markets
Speaker:Dr Miguel Figliozzi, University of Maryland
Abstract:Sequential auctions can be used to dynamically match shipments (shippers demand) and transportation capacity (carriers offer). They create new classes of on-line decision problems that arise in the operation of dynamic stochastic transportation systems. These decision problems include both dynamic vehicle routing and strategic bidding. Sequential auctions are used to study the impact of information availability on carriers behavior and learning as well as the performance of a wide range of dynamic vehicle routing technologies.
Date:21 October 2003
Topic:Truck or Train? A Stated Choice Study on Intermodalism in Argentina
Speakers:John Rose, ITLS-Sydney Research Analyst, Alejandra Efron, ITLS-Sydney PhD Student
Abstract:Intermodalism is regarded as the one of the most promising developments in freight transportation. The over-reliance on trucking has become critical in most countries around the world, and intermodal freight transportation involving both truck and rail is viewed as one way to reduce road congestion and alleviate associated environmental concerns. The decision process of shippers with regard to choice of transport mode related to freight movement is influenced by many different factors. In this paper we model this decision process. The study was conducted in Argentina during 2002 with the transportation modes investigated being road and rail within the context of freight movements. The methodology of the empirical study involved two stated preference (SP) choice experiments. Twelve design attributes were chosen to relate to dimensions of freight movements. For the first experiment, the alternatives available to the respondent were limited to the use of a single mode (truck only) or intermode (truck and train). For the second experiment, a third alternative was added allowing respondents to not use either of the aforementioned alternatives and wait for more favourable circumstances to arise. Results show that the attributes found to be statistically significant influences on the choice of selecting a transport mode when sending the freight is obligatory (first experiment) differ from those attributes that were found significant when the shipper is able to wait and send the freight later (second experiment). Several non design attributes were also found to be statistically significant contributors of choice process.
Date:14 October 2003
Topic:Defining and Measuring Transport System Capacity and Flexibility: Some Recent Research at the University of Pennsylvania Adobe PDF Document
Speaker:Professor Edward K. Morlok, UPS Foundation Professor of Transportation & Professor of Systems Engineering, University of Pennsylvania
Abstract:This presentation will summarize the results to date from a multi-year research program directed at defining and measuring transportation system capacity and flexibility. Flexibility and capacity are closely related. Interest in both concepts is growing, due to traffic increasing faster than expansion of network infrastructure, changes in supply chains and traffic patterns, and the concern for the vulnerability of the system to both natural disasters and terrorist actions. Models for measuring capacity and flexibility are described, along with results of pilot applications and an assessment of their practicality given data and other contextual considerations. This work has been supported by grants from the US National Science Foundation, the US Dept. of Transportation, the UPS Foundation Fund, Boeing and other private sector firms.
Date:1 October 2003
Topic:New Approaches to Traffic Monitoring by Floating Car Data Adobe PDF Document
Speaker:Dr Astrid Ghnemann, ITLS-Sydney Visiting Fellow, Institute of Transport Research at the German Aerospace Centre
Abstract:Intelligent traffic management is widely acknowledged as a means to optimise the utilisation of existing infrastructure capacities. A major requirement for intelligent traffic management is the collection of high quality data on traffic conditions in order to generate accurate real-time traffic information. Conventional data sources are based on traffic volume measurements, most often by inductive loops or infrared sensors. A shortcoming of these methods is the problem of reconstructing travel times and routes of vehicles in the road network. In our approach we generate this information by a large Floating-Car-Data (FCD) project in which we use data from taxi drivers in metropolitan areas. Based on this information, various applications such as real-time traffic monitoring, timedynamic routing and fleet management are established. In our paper we further discuss the additional benefit of this data base in building up dynamic emission models as a basis for integrating environmental aspects into intelligent traffic management systems.