Multi-Agent Architecture for Knowledge-Driven Decision Support - Distributed Artificial Intelligence, Agent Technology, and Collaborative Applications - page 39

focusing on providing intelligent distributed decision support that may be external to the company. In
the second case, the proposed architecture incorporates the W3C Web Services architecture that uses
the simple object access protocol (SOAP) and XML. The incorporation of this architecture creates a
flexible means of exposing the services of the agents using the Web Services architecture by a company
to its potential or existing global population of customers and suppliers.

iMPle Ment ation of the IKMdSA Ar ChITeCTure And ILLuSTr ATIVe
eXaMPle

The problem domain selected for the initial proof of concept is the play tennis decision problem (Mitch-
ell, 1997) using the ID3 decision tree method. The selection of the problem domain was due to it being
widely adopted (Mitchell, 1997) to represent decision problems in the ID3 decision tree research and
also for its simplicity in illustrating the proposed architecture. The decision problem for this problem
domain is to decide whether, or not, to play tennis on a particular day based on climatic conditions such
as the day's weather outlook, the level of humidity, the temperature, and the wind conditions. Figure
4 shows a schematic of the decision solutions under consideration. The leaf nodes of the decision tree
represent the final outcome of the decision of whether to play tennis on a certain day, based on what the
weather is like. The problem is simple to understand; yet it illustrates the fundamental requirements of
the system and provides an elegant way to test the various features of the agents and the architecture.
The end-user provides the existing weather condition to the user agent as input and the agent makes
a decision and presents the decision to the end-user whether or not tennis can be played that particular

Figure 4. Decision tree representation of the Play Tennis Problem (Adapted from Mitchell, 1997)

Outlook

Sunny Rain

Overcast

Humidity Wind

Yes

High Normal Strong Weak

Yes

No No Yes

((Outlook == Sunny) && (Humidity == Normal))
||
(Outlook == Overcast)
||
((Outlook == Rain) && (Wind == Weak))