Research on the Improved Combinatorial Prediction Model of Steel Price Based on Time Series

Accurately predicting the price change of steel (main building materials) is an effective means to control and manage the cost of construction projects. It is one of the ways for construction enterprises to reasonably allocate building materials, save resources, reduce carbon emissions and reduce environmental pollution. Based on the monthly historical price data of 100 steel rebar (16 mm) from November 2010 to February 2019, the separation and retrieval process of the four components in the time series are improved. The improved multiplicative and additive models were used to make separate predictions, and the reasonable weight is given to combine the multiplication and addition model by the reciprocal of variance method. Finally, an improved prediction model of steel bar price combination with higher prediction accuracy is obtained. The prediction results show that the improved multiplication model and addition model have higher prediction accuracy, their MAPE are 2.62% and 2.36% respectively. Moreover, the prediction accuracy of the combined model is even higher, its MAPE is 2.29%. The prediction accuracy of the improved composite model is higher than that of the individual models. The improved combined prediction model of reinforcement price based on time series method can provide some reference and help for cost control and management in construction engineering, further reduce resource waste and construction non-point source pollution