Data Mining and Knowledge Discovery in Industrial Engineering

A Special Issue of the Journal
Computers and Industrial Engineering
Vol. 43, No. 4, September 2002

Evangelos Triantaphyllou, T. Warren Liao, and S.S. Iyengar     Guest Editors

PREFACE (pages 657-660)



      It has often been said that we live in the "information age." This verdict is best manifested by the immense creation, availability, and use of humongous amounts of data. After expressing datasets in megabytes now we are expressing datasets in terms of gigabytes (please note that the term "mega" stands for "large" and the term "giga" stands for giant in Greek). At the same time, more and more frequently we started expressing datasets in terms of terabytes. It is not a coincidence that "teras" means monster in Greek! This proliferation of large amounts of data has created many new opportunities and also challenges. This is true in engineering, science, and business domains.
        The new filed of data mining (DM) and knowledge discovery from databases (KDD) has emerged as the new discipline in engineering and computer science to address the new opportunities and challenges. Some may claim that this is an old scientific field since always people wanted to be able to analyze vast amounts of data and extract useful information (new knowledge) from them. However, in the modern sense of DM and KDD the focus seems to be in extracting information that can be characterized as "knowledge" and the data can be very complex and in large amounts.
        Thus, it should not be a surprise that Industrial Engineering (IE) has been impacted by, and also influences, the new discipline of DM and KDD. The versatile domain of IE presents additional opportunities and challenges on its own right.
        This Special Issue aims at presenting some new theoretical results and also representative applications of the interface of DM and KDD with IE. Such an interface can be beneficial in two complementary ways. First, DM and KDD methods can be used to solve key IE problems. At the same time, however, IE problems can pose new challenges for developing new DM and KDD approaches. Therefore, this relationship is an interdisciplinary one and can benefit both DM/KDD and IE.
        The papers presented in this Special Issue with focus on DM and KDD in IE, are a small example of the significant potential that exists in interfacing the fields of DM/KDD and IE. It is impossible for a single journal issue to even scratch the surface of the entire spectrum of such developments. The three Guest Editors of this Special Issue believe that the papers presented here represent a selected anthology of some of the key developments in the interface of DM / KDD and IE.
        This Special Issue presents a total of 13 papers. The first seven papers are more of a theoretical nature, while the remaining six papers are more application oriented. The first paper is written by L.-Y. Zhai, L.-P. Khoo, and S.-C. Fok. These authors study the fundamental problem of how to extract the features that are pertinent to a DM/KDD problem. The authors propose an ingenious approach which is based on rough sets and genetic algorithms (GAs). The second paper is written by N. Ye and X. Li and discusses a new classification approach which is based on clustering. The potential of this approach is also studied on some benchmark datasets. A well- known approach for DM and KDD is to use neural networks (NNs). The third paper, written by P. Wu, presents a new search component for a NN approach that is based on fuzzy sets. The numerical results presented in this paper are very promising. The fourth paper is written by X. Huo and presents a rigorous statistical approach for the identification of embedded consecutive subsequences and their relation to some DM and spatial statistics problems.
        The fifth paper is written by G. Chen, Q. Wei, D. Liu, and G. Wets and presents a powerful approach for solving some problems by means of association rules. The power of this approach is based on its simplicity, which does not come at the sacrifice of its applicability and effectiveness. Most approaches that mine association rules from datasets, suffer of high complexity times. The new approach, called SAR for Simple Association Rules, seems to offer an efficient and effective alternative to some of the current methods. The sixth paper also deals with the subject of association rules. This paper is written by Y.Y. Hu, R.-S. Chen, and G.-H. Tzeng. The focus of this paper is on the mining of fuzzy association rules.
        The seventh paper is written by T.-L. Sun and W.-L. Kuo and presents a rather intriguing approach to DM and KDD. This approach uses some visual representations of the data. In this way, the role of the human analyst becomes a critical one since computerized systems cannot fully comprehend these visual representations. All the previous theoretical papers are also accompanied with small numerical explorations of the effectiveness of the proposed methods. The second half of this Special Issue is covered by six papers. These papers are mostly application oriented although they describe some new theoretical developments as well. The eighth paper is written by M. Kantardzic, B. Djulbegovic and H. Hamdan. It describes the application of DM and KDD to the medical problem of diagnosing polycythemia vera (PV). It also compares the new approach with the existing medical practice. The results are very interesting. The ninth paper presents an application of a new neural network (NN) model to weather forecasting. This paper is written by T.B. Trafalis, A. White, B. Santosa, and M.B. Richman. This application involves the processing of large and complex rainfall and weather datasets.
        The tenth paper describes an application of neural network methods in studying a macro- economic problem related to the assessment of the investment risks of a number of countries. The results of the NN approach are compared with results obtained by applying more traditional statistical approaches. The authors of this paper are I. Becerra-Fernandez, S.H. Zanakis, and S. Walczak.
        The eleventh paper is written by S.H. Ha, S.M. Bae, and S.C. Park. This paper presents the application of certain DM and KDD methods on discovering new marketing behaviors. This new knowledge is used next in designing new marketing strategies. The twelfth paper is written by D.E. Brown and J.R. Brence. It presents the application of some advanced DM and KDD methods to diagnosing corrosion problems on airplanes. The last paper is written by S. Morris, Z. Wu, C. DeYong, S. Salman, and D. Yemenu. This paper presents the development of a DM and KDD system for clustering and analyzing text documents. Visual approaches play a central role on these developments.
        The Guest Editors of this Special Issue are very thankful to all authors of these papers. Their patience and strive for excellence is highly appreciated. We hope that this Special Issue will become a forum for the dissemination of the current developments and also for initiating new partnerships among researchers and practitioners. This is why the Special Issue has a dedicated webpage with URL: http://cda4.imse.lsu.edu/books1/special_issue2/special2.htm Finally, the Guest Editors wish to express their gratitude, for his patience and guidance, to the Chief Editor of the Computers and IE journal; Dr. Mohamed I. Dessouky.

                                                Evangelos Triantaphyllou, T. Warren Liao,, and S. Sitharama Iyengar,     Guest Editors

                                                October 2001

Dr. Triantaphyllou's Homepage

Go back to the Special Issue's Homepage