Optimum bid markup calculation in competitive bidding environments using fuzzy artificial neural networks
Date
1998Google Scholar check
Keyword(s):
Metadata
Show full item recordAbstract
The study presented in this thesis seeks the development of a model to be used by the construction industry for optimum bid markup calculations in static and non-static competitive bidding environments. The methodology proposed, by which a competitive advantage in the competitive bidding market place might be gained by using advanced neural network technology, stems from the inadequacy of the currently available theoretical and analytical bidding models to capture the complexity of the competitive bidding problem in its entirety. This methodology aims for the provision of solutions to the problems posed by the complexity and non-linearity of competitive bidding, and faced by other currently available tools, and for the incorporation of both objective and subjective factors in the appraisal of the bidding strategy of a bidder and its competitors. The framework presented in this research work attempts to incorporate not only the quantitative but also the qualitative factors that are inherent in the prevalent competitive bidding environment, and to assess the uncertainties pertaining to bidding factors and the estimation of optimum markups. Even more importantly, the resultant model incorporates all historical project and competitor data and provides its users with the means to assess the probability of winning at suggested bid markups. The methodology developed utilizes available tools such as algorithms, database management systems, artificial intelligence techniques, fuzzy set theory, statistical analysis, and stochastic modeling to arrive at optimum bid markups. The multi-factored nature of bid markup decisions is accounted for by means of artificial neural networks, and fuzzy set theory is employed as an operant upon the subjective and qualitative factors utilized by artificial neural network models, to account for the fuzziness built in them. Central to the proposed bidding model are the probabilistic neural network and the method of Parzen Windows. The method of Parzen Windows, which is free of the stochastic independence or dependence assertions central to Friedman's and Gates's models respectively, is used for the estimation of the bid markup's underlying multi-dimensional probability distribution function, and for the calculation of the probability of winning and the optimum bid markup.