提升预测能力的方法研究

Forecasting of settlement is an important subject in the area of engineering field. The accurate forecasting of settlement is the ultimate goal and the core of forecasting. Besides, it is also the basis of guaranteeing the city and engineering security. Due to the impact of complicated and random dynamic change process of settlement, reliability and predictability of mathematical models always are limited in the settlement forecasting. In order to enhance the reliability and predictability of forecasting of mathematical model, using the settlement monitoring data of a high-speed railway roadbed in the loess region to establish the non-synchronization of the three exponential smoothing model, cross-certification analytic method and after-verification analytic method are used to compare and validate the fitting accuracy and forecasting accuracy on different synchronization models for getting the best reliable forecasting model; then, parallel-modification analytic method is used to modify forecasting results for enhancing forecasting capacity. The results showed that the cross-certification analytic method and after-verification analytic method can provide a guarantee for the best forecasting model, and parallel-modification analytic method can not only enhance forecasting accuracy and forecasting length, but improve the comprehensive forecasting capacity as well. At the same time, the three methods can be applied to other mathematical models, because of simple principle, easy operation and strong adaptability.