Study of construction bidding system based on combination of rough set theory and back-propagation network

The estimation of optimum markup percentage is a critical activity for a contractor to win the tender. It is affected by many factors. This paper presents a novel method of markup estimation combining rough sets (RS) theory and back-propagation (BP) network for construction project. RS theory is utilized as a preprocessor to delete the redundant irrelevant factors to the project markup. Then the relevant factors are used to train the BP network and predict the project markup. Actual prediction results show that the performance of RSBP model combing RS theory and BP model is superior to that of BP network with higher global convergence ability and higher computing speed. In addition, the mean relative error of RSBP model is also smaller than the BP model.