PSO-Optimized Grey-BPNN Hybrid Model for Predicting Construction Project Costs

Abstract

This paper combined the grey prediction model with a back-propagation neural network (BPNN) and optimized BPNN parameters using particle swarm optimization (PSO) to predict the cost of construction projects. Simulation experiments were conducted. In the experiment, the GM (1,1) was combined with a BPNN improved by PSO to predict the cost of construction projects. Forty construction project samples were used, among which 30 were included in the training set, and the remaining 10 were included in the test set. The influence of the number of BPNN hidden layer nodes and the type of activation function on the algorithm performance was tested in the experiment. Moreover, the model was compared with the grey prediction model GM (1,1) and the traditional BPNN model. It was found that the proposed prediction model was more accurate in predicting the construction project cost than the other two models. The mean absolute error (MAE) and root mean square error (RMSE) of the GM (1,1) model were 0.287 and 0.045, respectively. The MAE and RMSE of the traditional BPNN model were 0.113 and 0.020, respectively. The MAE and RMSE of the proposed prediction algorithm were 0.067 and 0.013, respectively