|
9:15 - 9:30 REGISTRATION
9:30
Discrete Ordered Median Problems: Dealing with Generalized Objective Functions in Facility Location
Natashia Boland, University of Melbourne,
reporting on joint work with: Patricia Dominguez-Marin, Stefan Nickel and Justin Puerto
Abstract:
Facility location problems are of critical importance in logistics. However there are a wide range of criteria commonly used to assess the quality of a solution. For example, the total distance of all customers from their nearest facility, or the maximum distance of any customer from its nearest facility, are both used. Combinations of these, and more complex functions, are often of interest in logistics applications. Recently, a facility location problem with a generalized objective function has been proposed, that captures each of these criteria as a special case. Known as the Discrete Ordered Median Problem, it is, however, very difficult to solve. In this talk the problem will be described, integer programming models for it discussed, and a specialized branch-and-bound method presented.
10:00
MODEL FOR A SPECULATIVE BUBBLE
B. D. Craven, University of Melbourne
Abstract
A mathematical model is proposed for a speculative stock market, relating to a boom and bust period, during which psychological factors (optimism) count for more than economic fundamentals. A MATLAB simulation shows that the model captures many of the qualitative features observed during such periods. (But it does not predict the date of the crash.)
10:30
Scheduling Jobs with Forbidden Zones
Amir Abdekhodaee and Andreas Ernst, CSIRO Mathematical and Information Sciences , Clayton, Vic 3169, Australia
Abstract
The problem of sequencing ships for a berth was considered where tidal constraints restrict the movement of ships in and out of the berth. However, there is no restriction on berth loading/unloading operations even when there is a low tide. The objective is to sequence the ships, with not necessarily identical processing times, so that the completion time of the last ship's operation can be minimised. We show that the problem is NP-hard and provide some numerical results based on an integer programming approach to the problem.
11:00-11:30 Morning Tea
11:30
Writing your own DP computer codes: a practitioner oriented guide
Moshe Sniedovich, Department of Mathematics and Statistics,
The University of Melbourne, Parkville, VIC 3052
Abstract:
Unlike the case of linear programming, there is no such thing as a "general purpose dynamic programming computer code". Consequently, in practice practitioners and researchersİ may have to write their own dynamic programming codes to suit their particular needs. There are clear indications that this is not always a straightforward task. In this presentation we shall discuss some of the fundamental practical aspects of developing dynamic programming codes for practical applications as well as research projects. The emphasis will be on key issues rather than details pertaining to specific computer platforms, problem instances and user interfaces. The discussion as a whole has been motivated by the speaker's experience over the last 35 years and is based on consulting and research activities in this area.
12:00
Rendezvous-evasion search in two boxes with incomplete information
S. Gal (University of Haifa) and J. V. Howard (London School of Economics)
Abstract:
An agent (who may or may not want to be found) is located in one of two boxes. At time 0 suppose that he is in box B. With probability p he wishes to be found, in which case he has been asked to stay in box B. With probability 1-p he tries to evade the searcher, in which case he may move between boxes A and B. The searcher looks into one of the boxes at times 1, 2, 3, .... . Between each search the agent may change boxes if he wants. The searcher is trying to minimise the expected time to discovery. The agent is trying to minimise this time if he wants to be found, but trying to maximise it otherwise. This paper finds a solution to this game (in a sense defined in the paper), associated strategies for the searcher and each type of agent, and a continuous value function v(p) giving the expected time for the agent to be discovered. The solution method (which is to solve an associated zero-sum game) would apply generally to this type of game of incomplete information.
12:30 - 1:30 Lunch
1:30
Predicting the Outcome of the 2003 Rugby World Cup
Stefan Yelas and Stephen Clarke, Swinburne University of Technology
Abstract:
A simple forecasting model was built in Excel to predict the results of each game and the tournament as a whole in the 2003 Rugby World Cup. An exponential smoothing technique was used to update the team ratings after each game, and optimised on all 566 games between the 20 World Cup teams from 1996. The model predicted the winning team, the margin of victory and the probability of a win. A tournament simulator used these predicted probabilities to calculate a team's chance of placing at any given time during the tournament. Match and tournament predictions have been regularly updated on our web site www.swin.edu.au/sport. The model has selected the correct winner of all 40 pool games, and the predicted margins have been used for profitable gambling.
2:00
Once upon a time there was exponential smoothing.
Ralph D Snyder, Department of Econometrics and Business Statistics, Monash University, Australia.
Abstract:
A revised version of the exponential smoothing method of forecasting is described. It is distinguished from its earlier incarnation by its reliance on sound statistical principles for maximum likelihood estimation, prediction and model selection. This new framework is contrasted with other methods of forecasting such as the Box Jenkins and state space approaches. It is argued that exponential smoothing is the most practical time series approach for tackling forecasting problems encountered in business, economics and finance.
2:30 Afternoon tea
Close
| 