Conceptual model for knowledge discovery process in databases based on multi-agent system
| DOI | https://doi.org/10.1108/VJIKMS-01-2015-0003 |
| Pages | 207-231 |
| Date | 09 May 2016 |
| Published date | 09 May 2016 |
| Author | Alireza Jahani,Peyman Akhavan,Mostafa Jafari,Mohammad Fathian |
| Subject Matter | Information & knowledge management,Knowledge management,Knowledge management systems |
Conceptual model for knowledge
discovery process in databases
based on multi-agent system
Alireza Jahani
Department of Information Technology, Mehralborz University,
Tehran, Iran
Peyman Akhavan
Department of Management, Malek Ashtar University of Technology,
Tehran, Iran, and
Mostafa Jafari and Mohammad Fathian
Industrial Engineering Department, Iran University of Science and
Technology, Tehran, Iran
Abstract
Purpose – Knowledge discovery in databases (KDD) is a tedious and repetitive process. A challenge for the
effective use of KDD is understanding and conrming its results derived from the harmonized process. To
exploit the advantages of agents’ application, this paper aims to propose a conceptual model based on a
multi-agent system (MAS) to control each step of the KDD process.
Design/methodology/approach – This paper reports the empirical ndings of a survey conducted
among academic and industrial sectors in Tehran, Iran. In this survey, the participants answered a
questionnaire about the main factors of designing a suitable model for the KDD process based on MAS. The
factor analysis reveals important insights of previous models developed by various researchers.
Findings – This research uses the survey results to nd six critical success factors, continuity in
renementand improvement; learning and acting concurrently; loosely or tightly coupled approach for using
technologies; cooperative, dynamic and exible environment; documentation and reporting; and extracting
and evaluating knowledge intelligently, for a proper conceptual model of the KDD process based on MAS.
Research limitations/implications – The proposed model reects all aspects of the KDD process by
applying the intelligent agents for each process steps. In addition, this research only considers the Iran
society; hence, it cannot be generalized to other nations, and it may need further research in other countries
and to be implemented in real-world business domains.
Originality/value – This research helps organizations to adopt a proposed model and implement a KDD
process to advantage the valuable knowledge that exists in their data resources.
Keywords Data mining, Factor analysis, Knowledge discovery process, Multi-agent systems,
Process models
Paper type Research paper
Introduction
Over the past decade, advances in data modeling approaches and database technologies
have led to better storage and maintenance of business data and knowledge (Fayyad
et al., 1996b). Businesses have become familiar with research areas such as knowledge
discovery in databases (KDD) and data mining (DM). Furthermore, the maturing KDD
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/2059-5891.htm
Knowledge
discovery
process
207
Received 12 January 2015
Revised 30 July 2015
14 November 2015
Accepted 7 December 2015
VINEJournal of Information and
KnowledgeManagement Systems
Vol.46 No. 2, 2016
pp.207-231
©Emerald Group Publishing Limited
2059-5891
DOI 10.1108/VJIKMS-01-2015-0003
technologies have been accepted increasingly, while human-driven aspects still need
further automation (Blake and Williams, 2003). KDD is used to address diverse and
often unprecedented questions on issues ranging from marketing to fraud detection, to
Web analysis and to command and control. Thus, one of the main challenges of
organizations are using data analysis and processing techniques appropriately and
correctly. The applicability of techniques depends on a number of factors, including the
question to be addressed, the characteristics of the data being studied and the history of
processing of those data (Jensen et al., 1999). Besides, the activities of the KDD teams
should be properly planned and coordinated.
Knowledge discovery is the nontrivial extraction of implicit, previously unknown
and potentially useful information from data (Frawley et al., 1992). Davies and Edwards
(1995) also believe that the KDD techniques are used to nd previously unknown
patterns in real-world data. The KDD process is a multiphase process involving several
steps like data preparation and preprocessing, searching for hypothesis generation,
pattern recognition, knowledge evaluation, representation, renement and management
(Fayyad et al., 1996a;Zhong and Ohsuga, 1995). Furthermore, the process may be
repeated at different stages when the database is updated. Besides, explicit
representation of the KDD processes has very important outcomes. First, effective KDD
requires managing dependencies between steps. Some steps may require, disallow or
enable other steps (Jensen et al., 1999). Second, the details of processes are essential to
determine the statistical validity of inductive inferences (Weiss and Kulikowski, 1991).
Third, the process details are necessary to validate the KDD results in a more general
way. Fourth, explicit representation of the KDD processes can help balance multiple
performance goals. Several approaches to a given analysis task may produce different
results of statistical validity, comprehensibility and ultimate utility (Jensen et al., 1999).
The paper is organized as follows. In the next section, relevant literature and related
works about the knowledge discovery process models and agent-based KDD process
models are reviewed. In the third section, the research methodology is presented and the
proposed process model is developed. A discussion about the results is presented in the
fourth section. Finally, the conclusions and recommendation for future works are
presented the last section.
Literature review
Although, the process-centric view has recently been widely accepted by researchers in
the KDD community as an important methodology for knowledge discovery, a key
shortcoming of the existing KDD processes is their reliance on human beings to plan and
control the KDD process steps (Zhong et al., 2002;Kargupta and Chan, 2000). The
suggested solution for addressing such problems is deploying agent-based systems in
the KDD processes. Agent-based systems belong to the most vibrant and important
areas of research and development that emerged in information technology since the
1990s (Luck et al., 2005). According to some researchers (Rudowsky, 2004;Wooldridge,
2009), intelligent agents are dened as agents capable of reactive, exible and
autonomous action to meet their design objectives. In some environments, various
agents work together to achieve multiple goals. Multi-agent system (MAS) is a system
composed of several agents, capable of mutual interaction in the form of message
passing, requesting, negotiating or producing changes in their common environment.
The usefulness of an agent’s features depends on their application. Therefore, it is
VJIKMS
46,2
208
To continue reading
Request your trial