A strategic decision support system for a consulting engineering firm using a hybrid neural network-expert system approach.

The vast majority of organisations operate in a dynamic and uncertain business environment. In order to remain competitive, these organisations need to think and act strategically, and many do so by employing the principles of strategic management. These can be applied equally well in the construction industry, where widespread changes in recent years have required professional consulting engineering firms to increasingly consider and develop strategic responses in order to compete successfully and effectively in their markets. It is argued that the effectiveness of the strategic decision making process can be improved with the use of a computer-based Strategic Decision Support System (SDSS). A SDSS is proposed that could - with appropriate adaptations - act as an intelligent assistant to a group of managers involved in making a strategic decision regarding the future of a professional consulting engineering firm. The proposed system was developed by integrating the two popular technologies of artificial neural networks (ANNs) and expert systems into a hybrid architecture.

The purpose of this thesis is to discuss the motivation for creating the proposed system and provide an overview of its development. After first introducing the topic, a literature review follows of the main concepts that were involved in the development of the proposed SDSS. Some of the existing SDSSs are next discussed, followed by an overview of the proposed system. The development of the system is then described in detail. This explains how the ANNs were trained and how genetic algorithms were also employed during this process. Furthermore, it explains how the expert system was built, and how this involved the creation of a suitable user interface, as well as a computerised version of the Delphi Technique for group decision making. The integration of the two components into a hybrid architecture is subsequently described. Finally, the discussion examines how validation of the proposed system was carried out, followed by a brief overview describing its use.

The thesis concludes by suggesting that the proposed SDSS is a useful and practical applications package which is believed to be the first of its kind. It contends that the creation and implementation of such a hybrid system is not only feasible but that is has considerable potential. The limitations of the proposed system are also outlined, followed by a series of recommendations for further research work.