论文部分内容阅读
为解决目前有关地下水动态变化研究方法中存在的诸多不足,采用非线性偏最小二乘回归(PLSR)分析方法。该方法首先对原自变量的每一维进行非线性变换,然后对因变量和新的自变量应用单因变量PLSR的简化算法进行回归求参,最后将回归系数通过逆变换代回到原来的变换式中,并最终求得原自变量和因变量的预测关系式。应用实例表明,建议方法将偏最小二乘方法和非线性元素有效地结合起来,可有效解决具有复杂非平稳动态特性的地下水动态水位预测问题,而且模型构建简单,计算简便,预测精度也有一定的提高。
In order to solve the shortcomings of current groundwater dynamic research methods, nonlinear partial least squares regression (PLSR) analysis was used. In this method, firstly, each dimension of the original independent variables is transformed nonlinearly, and then the simplified algorithm applying the one-way variable PLSR to the dependent variables and the new independent variables is used for regression. Finally, the regression coefficients are substituted back to the original In the transformation, and finally obtain the original independent variables and dependent variables of the predictive relationship. The application examples show that the proposed method effectively combines partial least square method with nonlinear elements and can effectively solve the dynamic water level prediction problem of groundwater with complex and non-stationary dynamic characteristics. Moreover, the proposed method has the advantages of simple construction, simple calculation and high prediction accuracy improve.