Chapter XIX: Data Mining in Designing an Agent-Based DSS

Chapter XIX - Data Mining in Designing an Agent-Based DSS
Data Mining: Opportunities and Challenges
by John Wang (ed)
Idea Group Publishing 2003

	Brought to you by Team-Fly

Christian B hm, GIDSATD UTN FRSF

ArgentinaMar a Rosa Galli, GIDSATD UTN FRSF and INGAR CONICET

ArgentinaOmar Chiotti, GIDSATD UTN FRSF and INGAR CONICET

Argentina

The aim of this work is to present a data-mining application to software engineering. Particularly, we describe the use of data mining in different parts of the design process of an agent-based architecture for a dynamic decision-support system.

The work is organized as follows: An introduction section defines the characteristics of a dynamic decision-support system and gives a brief background about the use of data mining and case-based reasoning in software engineering. A second section describes the use of data mining in designing the system knowledge bases. A third section presents the use of data mining in designing the learning process of the dynamic decision-support system. Finally, a fourth section describes the agent-based architecture we propose for the dynamic decision support system. It implements the mechanisms designed by using data mining to satisfy the system functionality.

INTRODUCTION

Enterprise management involves making many different decisions. An enterprise can be generally considered as being organized in several domains in which different types of activities are performed and decisions are made. These decisions are associated to their corresponding activities but are usually closely related to decisions already made in other domains of the enterprise; that is, the different decision points of an organization often need information to be available in other enterprise domains. The enterprise integration systems provide the automatic communication among these points by transferring previously specified data (Shen & Norrie, 1999). These systems are designed based on the habitual requirements of each domain. But there exist non-foreseen situations that require information exchange. This occurs whenever new or non-habitual decisions must be made. In these cases, it is necessary to use a dynamic Decision-Support System (DSS) (Cabral et al., 2000).

A dynamic Decision Support System is a system able to look for information, analyzing where it is available or can be generated.

A dynamic DSS establishes contact among domains for the acquisition of necessary information for decision making. This implies that the information to be transferred and the domains to be communicated with are not specified on the system design time. The DSS itself should interpret the information requirement and infer which domain can answer it on the run time.

A dynamic DSS must operate in the following way: When a user of the system needs some information, he/she makes a query in natural language, and the dynamic DSS transfers that information requirement to a domain that can satisfy it. For that purpose, the system determines the sites that offer a greater possibility of providing the required information and those are targeted first. This defines the main functionality of this information system. Then, when it gets the answer from a domain, it takes this information to the domain that asked for it.

To develop a system with the aforementioned functionality, it must be provided with: (1) knowledge about the information that can be provided by domains; and (2) the capacity for updating that knowledge, learning from the cases that result from its operation. This chapter will present a description of how data-mining techniques have been used for designing a system with the aforementioned characteristics.

Dingsoyr (1998) presents several possible uses of data mining and case-based reasoning. He classifies these uses into two groups: data mining in case-based reasoning, and case-based reasoning in data mining. He sketches the following six possible uses:

1. Data mining search is the case. That is, the information about the search results and the whole knowledge discovery in database process might be stored in a case so that extra time will not be spent on mining the same information more than once. Rodriguez, Ramos, and Henriques (2000) presented a heterogeneous architecture for knowledge extraction from data. The architecture combines the knowledge discovery in database process with a case-based reasoning system intended to record the successful knowledge extraction experiments to be shared later.

2. Case-based reasoning provides information that can be used in providing some background knowledge about features in a database, for instance, the weight of features for a classifier can be learned.

3. Find features from cases base for a case. It might be valuable for classifying the cases in the cases base. This might speed up the computation of similarity metrics in the retrieve step.

4. Find features from a database for a case to supplement the information given in it.

5. Find domain knowledge by mining a database or a cases base in the form of functions, rules, or causal graphs that can be used later to identify features and explain away unimportant features.

6. Construct artificial cases from a database that are not present in a cases base. This would require a tight integration where the data-mining algorithm would search for patterns in the form of cases that could be evaluated by a novelty function that gives high values to cases not present in the base.

Ruiz, Fisler, and Cole (2001) describe a project of data-mining application on software engineering. Specifically, they propose to simulate the designed system to generate cases. Such cases are mined to detect possible behavioral patterns of the design.

It is necessary to highlight that data mining and case-based reasoning applications on system design offer a still little developed potential. On the other hand, in designing an information system, it is possible that several of the alternatives sketched by Dingsoyr (1998) are simultaneously used. In this work, we describe the use of data mining in different parts of the design process of a dynamic decision-support system architecture. By using data mining, we could define the structure of the system knowledge base, the data structure of the system cases base, and the rules of relationships among cases that were used to design the system learning mechanism.

	Brought to you by Team-Fly