A neurofuzzy expert system for competitive tendering in civil engineering

Competitive tendering is one of the most critical activities of contractors in the construction industry. A contractor must first decide whether to bid or not for a new project. If the bid decision was made, a cost estimate needs to be produced considering uncertainties involved in pricing the required materials, plant and labour and, thereafter, a mark up should be determined and added to the cost estimate as a coverage of profit and an allowance for unexpected risks. The tendering decisions; i.e. whether or not to bid and how much to mark up the estimated cost, are very important as they have profound effects on the day-to-day operations and the long-term performance of the construction firm. The importance of these two decisions lies in the fact that the success of a construction organisation is dependent on their outcomes. Additionally, these decisions are very complex because they are liable to be affected by many internal and external factors. In practice, however, the bidding decisions are usually made in a largely subjective manner. The absence of a suitable structured basis often results in mistakes causing loss to contractors and adversely affecting the industry. The main objective of the present study is to develop a simple-to-use tendering strategy model for possible implementation in the Syrian construction industry. During the last fifty years, many attempts have been made to model the process of making the bidding decisions. The majority of the developed models were based of the probability and the utility theories. The mathematical complexity of these modes, their over-simplified assumptions, and the necessity of historical data made them inapplicable in the construction industry. Other models were developed using regression analysis and multi-criteria decision analysis techniques. These models have many advantages over the probability and utility models. For example, they represent the bidding process more realistically as they account for multiple factors that affect this process. Also, the expert systems and the artificial neural network techniques were applied to the bidding process and helped to achieve some improvement over previous models. More recently, many researchers have proposed bidding strategies based on the fuzzy set theory. They claimed that fuzzy set theory is very suitable for the subjective nature of the tendering decisions. However, there is not a strong agreement among researchers on which modelling technique is the best for developing practical and more applicable tendering models. Therefore, based on the literature review, the modelling techniques that proved to be useful in previous studies were selected and used in the current work. These are regression analysis, decision analysis, and the artificial neural network techniques. The neural networks model was more reliable compared with the other models. Attempting to achieve more improvements, a new technology called neurofuzzy was implemented. This technology is a combination of neural networks and fuzzy expert systems. The application of this powerful tool has enabled an innovative tendering strategy model to be developed. This model has numerous advantages over all previous bidding models. It was implemented in a user-friendly computer prototype called NET (Neurofuzzy Expert systems for competitive Tendering in civil engineering). Testing NET on real life bidding situations provided evidence that it could be used in practice with great confidence. Unlike most previous bidding strategy models, NET can provide guidance in making the bid/no bid decision and in setting a suitable mark up size. This model provides civil engineering contractors with a standard methodology to improve the quality of their tendering decisions. It does not require any historical data about previous projects or potential competitors. Also, the user does not need to perform any mathematical computations. All he/she needs is to provide his/her subjective assessments of the bidding situation under consideration. In addition to all these advantages, the proposed model can be modified very easily to suit certain tendering policies by learning from new examples, adding new rules to the knowledge base, removing existing rules or fine-tuning their associative importance.