A Data Mining Framework for Training Intelligent Agents
The main goal of the Agent Academy project is to develop a framework
for enhancing agents' functionality and intelligence through the use of
Data Mining techniques. This particular framework is concerned with two
major issues: the first one focuses on the way the AA trains the agents,
in order to provide them with the necessary skills to successfully accomplish
their assigned tasks. This mechanism will combine the use of DM techniques
to improve the decision making process in IAs, the training of the newly
"born" agents, as well as the way the AA deals with different
types of agents. The second major issue involves techniques by which the
AA obtains information on the way it should train its own agents.
Figure 1 illustrates the basic modules of AA and the corresponding packages. A request for the creation of a new agent comes from a user specifying the application type and a set of requirements. With the help of the Data Miner the appropriate agent type is selected from a library of generic agents and is sent to the Agent Training module. After the training is complete, the now intelligent agent, armed with tools for reporting its behavior to the AA, is released to the world. The AA continuously receives feedback from mobile agents roaming the web and updates its agent use repository.
Figure 1. Basic modules of AA and the corresponding packages.
Work is structured in a set of seven technical WorkPackages (WPs). Figure
2 depicts the project outline, which provides a convenient representation
of project flow control in regard to the specific WPs. We begin with the
requirements analysis and specification of Agent Academy and then we proceed
to the development of the basic platform. This platform will be tested
for technical correctness but it will also evaluated against the three
test case scenarios that will be developed towards the last half of the
project. There will be an evaluation of the test case suite followed by
the overall assessment of Agent Academy. We will also develop a business
In the following (figure 2) we present a general description of each workpackage.
Figure 2. Agent Academy outline.
In the following table Agent Academy Workpackages are presented.
Work in WP1 includes all activities and decisions that
will result in the conceptual and functional definition of the multi -
agent framework generic model, the Agent Academy. The model will include
all the operational guidelines and instructions for the creation, coordination
and management of the final framework architecture.
Work in WP 2 will focus on the development of the AA tools that will enable the collection of generic mobile agents behavioral characteristics over the Web and the maintenance and updating of a data repository with relevant information. All collected data will guide the DM procedure on agent experience and will facilitate the generation of required information that will enhance agent intelligence. This work package includes the development of associated behavior tracking tools for agents trained by AA and the development of tools for reporting agent activities in remote web servers. It also aims to improve agent interoperability based on existing specifications.
WP 3 - Data Mining techniques for Agent intelligence improvement
The main objectives of WP 3 are to exploit DM techniques for improving the decision making process and to develop new or refine existing DM techniques for Intelligent Agent use and behavior. Work in WP 3 will focus on the development of the Data Miner component, which will provide the link between information collection and agent training and the Agent Factory component.
WP 4 - Integration and initial evaluation of the Agent Academy platform
WP 4 aims to integrate the core software components of
AA developed in the previous software packages into the basic software
platform. The integrated framework will be evaluated for quality assurance
and agreement to the initially designed requirements and specifications.
Work in WP 4 is focused on the integration of the Agent Training Module,
the integration of the main software platform and the functional evaluation
of the final product. Work in WP 4 is focused on the integration of the
Agent Training Module, the integration of the main software platform and
the functional evaluation of the final product.
The objective of WP 5 is to demonstrate the use of AA in industrial applications and assess its benefits. WP 5 will deploy three test case scenarios in real SME environments, apply the AA platform and measure the impact.
WP 6 - Dissemination and industrial implementation
Work in WP 6 involves the development of an exploitation plan for the results of the project. This covers the commercialization of the developed methods and software as well as the dissemination of the created know-how in this field. According to the different natures of the participants, different dissemination and exploitation strategies will be developed. Finally, work in this WP will be targeted towards the active contribution to European and international standardization, regulation and policy making activities. WP6 actions will be conducted in parallel to the design, implementation, and evaluation phases of the project.
This WP includes management, reporting, quality assurance, etc. The project coordination belongs to CERTH that will also be the liaison between the Consortium and the EC (through the Project Officer).