本文在理论上从介绍经典的多元线性回归开始，构架了统计学回归模型的基础。并给出了多元回归模型的显著性检验方法、优良程度的评判标准，以及模型中的问题与改进。面对高维数据，顺势提出了降维思想，Lasso回归是本文介绍的重点。本文介绍了Lasso回归的定义以及性质，还有方差的估计，该法可以有效解决最小二乘法和岭回归不能解决的问题，适用于样本容量小于自变量个数的情况。

Theoretically, this paper begins with an introduction to classical multiple linear regression and constructs the basis of a statistical regression model. The significance test method of the multiple regression model, the criteria for judging the degree of excellence, and the problems and improvements in the model are also given. In the face of high-dimensional data, homeopathy puts forward the idea of dimensionality reduction, and Lasso regression is the focus of this article. This paper introduces the definition and properties of Lasso regression, as well as the estimation of variance, which can effectively solve the problem that least squares and ridge regression cannot solve, and is suitable for cases where the sample size is less than the number of independent variables.