Information Technology Reference
In-Depth Information
considering its history. Such knowledge is used in the BL to decide when to bid for flight
tickets, and in this layer, to calculate the minimum asking prices when a decreasing
trend is identified. Finally, for the entertainment auctions, the agent has the same KL as
the RB traders, described in Section 4.
5.3
The Information Layer
Having obtained the private information about its client preferences, the agent
i
then
extracts all market information it requires in order to build the knowledge used in its
strategy. Indeed, it tracks information relevant to the TAC Travel Game, such as the
running time of the game and which auctions have closed (which are described by
H
(
p
M
(
t
k−
1
))), as well as the clients' preferences that do not change during the game
(which are described by
H
(
p
i
(
t
k−
1
))). When it considers the individual auctions, the
agent has to record the history of published information (bids and asks where available).
In the flight auctions, the history of flight prices is required to estimate the trend, which
represents vital knowledge. In the hotel auctions, the history of the publicly announced
16th highest price can be recorded up to when the auction closes. Such information can
be used to estimate the clearing price of the hotel auctions in future TAC games. Finally,
for the entertainment auctions, the agent has the same IL as the RB traders.
6
Conclusions and Future Work
As electronic marketplaces are being used on a broader scale, we believe software
agents will increasingly dominate the trading landscape. Their ability to make informed
decisions, based on the plenitude of market information, to a degree that human traders
can never achieve, make them ideal candidates for traders. However, as this new breed
of agents are populating the markets, it is becoming a fundamental challenge to design
strategies that can efficiently harness the avalanche of information that is available into
efficient trading behaviour. Given this, the objective of this paper is to provide a sys-
tematic framework for designing such strategies. To this end, we proposed a framework
that can be broken down into three principal components; namely the behavioural layer,
the knowledge layer and the information layer. In so doing, we believe this work is an
important preliminary step towards guiding the strategy designer by identifying the key
models and concepts that are relevant to this task. We applied this model to analyse
a selection of strategies in the CDA mechanism and showed its use when designing a
novel strategy for the TAC Travel Game. Our approach allowed us to first decide upon
the general outline of the strategic behaviour of the TAC strategy, and then delve into
the complex task of implementing it. For the future, we obviously need to verify our
framework further by applying it to different types of market institutions.
Acknowledgements
The authors would like to thank anonymous reviewers for helpful comments. Perukrish-
nen Vytelingum is funded by the DIF-DTC 8.6 project (www.difdtc.com) and Rajdeep
K. Dash is funded by a BAE Systems studentship.
