偏最小二乘回归法利用了成分提取的工作思路,在变量系统中选取若干对系统具有最佳解释能力的新综合变量,而非直接考虑因变量与自变量进行回归建模,因而能有效地解决自变量之间具有多重相关性的问题。应用该法针对边坡工程安全监测变量及其影响因子间的复杂相关性进行研究。通过对锦屏一级水电站左岸边坡深部变形的实测资料的建模分析表明,偏最小二乘回归法适用于对复杂相关的因子进行建模,通过交叉有效性检验确保模型的精度,同时对引起边坡变形的影响因子的重要程度进行排序,因而在边坡工程安全监测资料的统计分析方向具有广阔的应用前景。 Partial Least Squares Regression uses the working principle of component extraction, selects a number of new synthetic variables that have the best explanatory power in the variable system instead of directly considering the dependent variables and independent variables regression modeling, which can effectively Solve the issue of multiple correlations between independent variables. This method is applied to study the complex correlation between safety monitoring variables of slope engineering and its influencing factors. Through the modeling analysis of the measured data of the deep deformation of the left bank slope of Jinping I Hydropower Station, it shows that PLS method is suitable for modeling the complicated correlative factors and ensuring the accuracy of the model by cross-validation. The importance of influencing factors of slope deformation is ranked, so it has a broad application prospect in the statistical analysis direction of slope engineering safety monitoring data.