A fuzzy neural network (FNN) model embedded with a modified particle swarm optimizer (MPSO) is proposed in this study for cost estimation of construction projects. The proposed method has advantages over traditional FNN approaches in ways of generalization ability and automatic parameter optimization. Comparative studies show that this improved model is also superior to those of BP (back propagation) neural network, PSO-BP and PSO-FNN. Effectively improved accuracy of the developed model in cost estimation of construction project was discussed in case studies. This study provides a base for decision-making in the management of construction project.