Forecasting construction industry-level total factor productivity growth using neural network modeling

Total Factor Productivity (TFP) is widely recognised as a better indicator than Labour Productivity and Multi-Factor Productivity to represent industry-level productivity performance. Productivity is the key determinant of a nation's standard of living and an industry's competitiveness. As such, the ability to predict trends in TFP growth in the construction industry is very important. The factors influencing TFP growth in the construction industry are complicatedly interrelated. This fact made the conventional regression method highly inadaptable to such complex multi-attribute nonlinear mappings. As an AI information-processing tool, the artificial neural network (ANN) system has been proven to be a powerful approach to solving complex nonlinear mappings with higher accuracy than regression methods. However, so far, there has been little application of ANNs in predicting TFP growth in the construction field. This study will for the first time, apply the concepts of ANNs to develop a model to forecast the TFP growth in the case of the construction industry of Singapore. Macro-level information processing models are useful in monitoring and predicting the performance of the construction industry as a whole. With the need to manage construction performance information at all three levels, namely, industry, firm and site, this study looks specifically at developing an 'intelligent' model for forecasting industry-level productivity. Meanwhile, using the same set of data, a model developed by the Multiple Linear Regression method will serve as a benchmark to judge the performance of the ANN model. The ANN model, compared with the traditional regression model, would be expected to have better forecasting ability for TFP growth in the construction industry, in terms of accuracy.