Forecast of Safety Situation in Construction Industry based on Composite Model

Forecast of production safety situation in construction industry is a complicated non-linear problem, whose evolutional process has apparent randomness and volatility. According to the data of production safety situation in construction industry between 2010 and 2011, the back-propagation neural network model, the moving average model and the exponential smoothing model are adopted to predict, respectively. Combined with the characteristics of the three forecast models, a new forecast model with non-negative weights is proposed. Comparison with practical situation indicates that the proposed forecast model can overcome the shortcomings of the single forecast model and solve the forecast difficulties caused by safety indicators under random system states. The studied results shows that the proposed composite model with non-negative weights is feasible for the forecast of production safety situation in construction industry.