Bid mark-up selection using artificial neural networks and an entropy metric

Abstract

The purpose of the paper is to perform bid mark-up optimisation through the use of artificial neural networks (ANN) and a metric of the selected bid mark-up's derived entropy. The scope is to provide an alternative, entropy-based method for bid mark-up optimisation that improves on the analytical models of Friedman and Gates. The proposed method enables the incorporation of bid parameters through the use of ANN's pattern recognition capabilities and the integration of these parameters with a mark-up selection process that relies on the entropy produced by possible mark-up values. The entropy metric used is the product of the probability of winning over the bidder's competitors multiplied by the natural logarithm of the inverse of this probability. The case study results show that the proposed entropy-based bidding model compares favourably with the prevailing competitive bidding models of Friedman and Gates, resulting in higher optimisation with regards to the number of jobs won, the monetary value of contracts awarded and the value of "money left on the table". Furthermore, the method allows for the incorporation of several objective and subjective bid parameters, in contrast to Friedman's and Gates's models, which are based solely on the bid mark-up history of a bidder's competitors. While the proposed method is a useful tool for the selection of optimal bid mark-up values, it requires historical data on the bidding behaviour of key competitors, much like the classic bidding models of Friedman and Gates. The method is suitable for quantifying objective and subjective competitive bidding parameters and for optimising bid mark-up values.