Data Mining-Driven Analysis and Decomposition in Agent Supply Chain Management Networks

Only available on StudyMode
  • Download(s) : 17
  • Published : October 4, 2012
Open Document
Text Preview
Data Mining-Driven Analysis and Decomposition in Agent Supply Chain Management Networks
Kyriakos C. Chatzidimitriou1, Andreas L. Symeonidis1,2 and Pericles A. Mitkas1,2 1
Department of Electrical and Computer Engineering
Aristotle University of Thessaloniki, GR541 24, Thessaloniki, Greece 2
Intelligent Systems and Software Engineering Laboratory
Informatics and Telematics Institute/CERTH, 57001, Thessaloniki, Greece {kyrcha,asymeon}@issel.ee.auth.gr, mitkas@auth.gr

Abstract
In complex and dynamic environments where interdependencies cannot monotonously determine causality, data mining techniques may be employed in order to analyze the
problem, extract key features and identify pivotal factors.
Typical cases of such complexity and dynamicity are supply chain networks, where a number of involved stakeholders struggle towards their own benefit. These stakeholders may be agents with varying degrees of autonomy and intelligence, in a constant effort to establish beneficiary contracts and maximize own revenue. In this paper, we illustrate the

benefits of data mining analysis on a well-established agent supply chain management network. We apply data mining
techniques, both at a macro and micro level, analyze the results and discuss them in the context of agent performance improvement.

1. Introduction
As agent technology matures in time, autonomous agents
gain applicability and trust in trading and auctioning goods in real-world electronic markets, as well as in managing more complex environments like supply chain networks [4, 8]. Various approaches are followed in order to determine the optimal agent strategy with respect to the

challenges they (agents) come up against. The plethora of
data generated by these highly dynamic markets can be exploited in various contexts, either for online, or for a posteriori analysis. Our work attempts to evaluate data mining (DM) methodologies for analyzing and improving agent behavior based on market data analysis for an agent supply chain testbed, the Trading Agent Competition, Supply Chain Management game (TAC SCM) [1]. TAC SCM is

a dynamic, stochastic, partially observable and competitive

multi-agent environment that simulates an instantiation of a realistic supply chain management scenario. It allows for a
huge strategy space for agents to apply, while it is very challenging in the sense that decisions taken on any given day may affect the state of the agent many days later. Due to the time constraints and the huge dimensionality of the game,

analytic strategies and algorithms are much more difficult
to apply and succeed than in simpler auction settings. All
the above make the TAC SCM game a very good testbed
for applying DM techniques and establishing their worth
as intelligent agent modules against various conditions and
opponents.

In general, DM analysis of a domain may identify opportunities and help the agent designer gain a predictive edge over his/her opponents. In this context, we have applied DM in order to: a) provide a DM-driven performance evaluation of the agents over the past competitions, derive

conclusions as to what makes an agent successful, and subsequently guide the design of SCM agents and, b) deduce models of market behavior in order to build a robust and
high-performing agent that will be able to be ahead of its
competition. The resulting agent was used as a strawman
against other methodologies that involve on-line learning
and/or heuristic approaches. In this work, we focus our DM
methodology on the bidding component of the agent, which
is the one that comes in direct competition with its counterparts.

The rest of the paper is organized as follows: Section 2
provides an overview of the TAC SCM and references related work on the domain. Section 3 describes performance evaluation models, while Section 4 presents the bidding algorithm. Finally, Section 5 summarizes work conducted.

2. Background & Related Work
Within the...
tracking img