偏最小二乘回归与灰色模型耦合预测城市用水量

Coupling Partial Least-Squares Regression with Grey Model(1,1) to Forecaste Urban Water Consumption

影响城市用水量的各个因素,存在多重相关性,采用传统最小二乘回归法建模,其估计参数存在较大误差,预测精度降低。运用偏最小二乘回归法建立城市用水量的预测模型可以克服变量间的多重相关性影响,并可以很好地解释因变量;采用GM(1,1)建立的城市用水量预测模型,能够克服参数的非线性干扰,进行中长期预测。结果和实际符合,将两者进行耦合,充分利用了两种模型的优点,预测结果更为合理可靠。

There is multi-correlation among the factors which affect the city water consumption. The result obtained by the traditional least sequare method has a greater error than the true value. The partial least-square regression （PLS） is applied to set up the urban water use model based on the main component analysis and typical correlation analysis, it can solve the problem of interactive correlation among the independent variables and explain the dependent variables very well. The GM（ 1,1 ） model which is adopted to build the urban water consumption can overcome the obstruction of none-linear parameters. The results and the practical data coincide very well in the mid-long forecast. The coupling of the PLS and GM（ 1,1 ） model can obtain a good result which is more reasonable and reliable.