International Trading Agent Competition. In.pdf

SouthamptonTAC: Designing a Successful Trading Agent Minghua He and Nicholas R. Jennings 1 Abstract. This paper describes the design, implementation and evaluation of SouthamptonTAC, an agent that participated in the Sec- ond International Trading Agent Competition. In the course of the competition’s approximately 600 games, SouthamptonTAC achieved the highest mean score and the lowest standard deviation. 1 INTRODUCTION The Second International Trading Agent Competition [3] (TAC) in- volved 27 agents developed by universities and research labs from around the world. In the course of the competition, some 600 games were played and our agent (SouthamptonTAC) registered the high- est mean score and the lowest standard deviation. In a TAC trad- ing game, there are 8 software agents (entrants to the competition) that compete against each other in a variety of auctions to assemble travel packages for their individual customers according to their pref- erences for the trip. A valid travel package for an individual customer consists of (i) a round trip flight during a 5-day period (between TAC- town and Tampa) and (ii) a stay at the same hotel 2 for every night between their arrival and departure dates. Moreover, arranging ap- propriate entertainment events during the trip increases the utility for the customers. The objective of each agent is to maximise the total satisfaction of its 8 customers (i.e., the sum of the customers’ util- ities). Customers have individual preferences over which days they want to be in Tampa, the type of hotel they stay in, and which enter- tainment they want to attend. This data is randomly generated by the TAC server in each game (see Table 1 for an example). Table 1. SouthamptonTAC’s customer preferences for game 5722. PAD and PDD stand for preferred arrival and preferred departure date. HV stands for the reservation value of staying in the Tampa Tower hotel, and WV, PV and MV stand for the utility associated with attending Alligator Wrestlin