Abstract: All primitive variables are contained in partial least－squares regression model. When there are lots of variables, the result is extremely complicated, that result analysis and explanation is very difficult. Through model lamination establishment, the problem is solved by Hierarchical PLS (Hi-PLS) regression. Application in engineering projects indicates that the model precision is higher and the model is suitable for regression analysis which has large-amount variables.

摘要： 偏最小二乘回归模型中包含所有原始选择的变量，当自变量较多时，因得到的模型结果十分庞杂而难以分析和解释。本文采用递阶偏最小二乘（Hierarchical PLS, Hi-PLS）回归方法，通过分层建立模型的方法有效解决了这一问题。工程实践表明，本模型精度较高，特别适用于大规模变量集合的回归分析。