A neural network model for decision making With application in construction management

Journal of International Information Management Volume 3 Issue 2 Article 3 1994 A neural network model for decision making With application in construction management Mirza B. Murtaza Prairie View A & M University Debroah J. Fisher University of Houston Follow this and additional works at: https://scholarworks.lib.csusb.edu/jiim Part of the Management Information Systems Commons Recommended Citation Murtaza, Mirza B. and Fisher, Debroah J. (1994) "A neural network model for decision making With application in construction management," Journal of International Information Management: Vol. 3 : Iss. 2 , Article 3. Available at: https://scholarworks.lib.csusb.edu/jiim/vol3/iss2/3 This Article is brought to you for free and open access by CSUSB ScholarWorks. It has been accepted for inclusion in Journal of International Information Management by an authorized editor of CSUSB ScholarWorks. For more information, please contact . Murtaza and Fisher: A neural network model for decision making With application in co A Neural Network Journal of International Information Management A neural network model for decision making With application in construction management Mirza B. Murtaza Prairie View A & M University Deborah J. Fisher University of Houston ABSTRACT In this paper, an innovative approach is presented to decision making using self-organiz ing multi-layered neural networks. The model helps make a decision whether to use a conven tional stick-built method or to use some degree of modularization when building an industrial process plant - a problem considered very important in construction management because of its economic impact. The objective of this paper is to show that both expert system and neural network approaches can be useful for decision making problems. However, in some situations a neural network approach can outperform the expert system approach. A brief overview of prior approach to modular construction decision making is provided in this paper and the reasons for using a neural network approach are also discussed. The architecture, knowledge representation, and training procedure for the neural network para digms used are described. The performance of the trained neural network system and its com parison with the recommendations provided by human experts and the expert system are also presented. INTRODUCTION This paper deals with the design and development of a neural network based decision making model. The model helps management personnel in the construction industry decide whether to use a conventional construction method or to use certain degree of modular construc tion method when planning to build an industrial process plant either within or outside the United States. The feasibility of construction modularization depends on the specific project situation, organizations involved, social, legal and environmental conditions. In some obvious project environments, such as remote sites, harsh weather conditions, etc., modularization rep resents the only feasible choice. On the other hand, in some other situations the decision to modularize is not as obvious. Therefore, at the initial stage of a project, management must decide whether to investigate the modularization potential. 27 Published by CSUSB ScholarWorks, 1994 1 Journal of International Information Management, Vol. 3 [1994], Iss. 2, Art. 3 Journal of International Information Management Volume 3, Number 2 In the past, there has never been an easy-to-use method that can be used to determine modularization feasibility. The only way in which companies have utilized modularization in the past has been when an expert in the field was consulted from within the organization or from another organization. However, there are many engineering and construction and owner compa nies which need to be able to determine modularization feasibility of a project in a simpler and more easily accessible way. An expert system has already been designed to achieve this goal (Murtaza, Fisher and Skibniewski, 1993). However, the results of the research presented here show that a neural network approach outperforms an expert system for the present problem. Additionally, the neural network based system can handle the inexact and incomplete inputs in order to reach a conclusion (Kamarthi, Sanvido & Kurmara, 1992) and, thus, is more appropri ate for unstructured decision making environments like construction modularization. The next section of this paper presents a brief overview of expert system approach devel oped previously for modular construction decision making. The paper then discusses the archi tecture of the neural network paradigms used, and also describes the overall architecture and training procedure of the neural network system. The performance of the neural network system and its validation results are also provided. EXPERT SYSTEM APPROACH Modular construction is a method for constructing units of a project in a remote location from the final project site. Modularization brings the advantage of the manufacturing process to the construction industry, such as a controlled environment (temperature and lighting), improved quality control, improved safety, etc. Modularization offers an opportunity to improve a variety of performance parameters relating to the project, such as cost and schedule. A module is a remotely assembled unit. It is usually the largest transportable unit or component of a facility. It has all structural elements, finishes, and process components fitted. Modules may contain pre fabricated components or preassemblies. An expert system for construction modularization decision making has already been de veloped (Murtaza, Fisher & Skibniewski, 1993). During the knowledge-base development phase, several hours of knowledge acquisition sessions were held with the experts at the major engi neering and construction, fabrication and owner firms in the construction industry. These ses sions with modularization experts provided an extensive amount of information about the modularization feasibility study process. The most important discovery was the determination of factors to consider when such a study is performed. These factors can be categorized in the following five groups: Plant Location, Labor Considerations, Environmental and Organizational Factors, Plant Characteristics, and Project Risks (Murtaza, 1993). The analysis of project location includes such factors as accessibility, climatic conditions, bulk commodity quality and availability, construction equipment quality and availability, trans portation mode, transport equipment availability and timing. Labor skills, productivity and type (union or non-union) are some of the factors included in the labor related category. Some of the factors related to the project characteristics are repeatability, proprietary security, project type https://scholarworks.lib.csusb.edu/jiim/vol3/iss2/3 28 2 Murtaza and Fisher: A neural network m (...truncated)