不等时距GM(1,1)模型预测地基沉降研究

分别采用直线插值、三次样条插值、BP神经网络3种方法,用M ATLAB语言编制程序将不等时距序列转化为等时距序列,采用灰色理论预测沉降.由于BP神经网络强大的非线性映射功能,可以避免常规插值法所造成的一系列误差.实际工程应用结果表明,利用直线插值、三次样条插值和BP神经网络与灰色理论联合建模所得的预测值与实测值的最大相对误差分别为17.2%,5.9%和4.6%.由此可见BP神经网络和灰色理论联合建立的GM(1,1)模型用于预测路基沉降最为精确.

Application of BP Neural Networks in Unequal Time Interval Model of GM(1,1)

The straight line interpolation,the cubic spline interpolation,and BP neural networks are used respectively to change the unequal time interval data sequence into an equal time interval data sequence within MATLAB language generator,and the settlement is forecasted according to gray theory.Because of the powerful nonlinear mapping function of BP neural networks,a series of error brought about by general interpolation law can be avoided.The results of actual project indicate that the maximal proportional error between the forecast values,gained by making use of the straight line interpolation gray theory model,the cubic spline interpolation gray theory model,and BP neural networks gray theory model,and the observation data are 17.2%,5.9% and 4.6% respectively.It is concluded that the BP neural network grey theory model is the most precise one to forecast the settlement.