LASSO回归

LASSO回归总结

LASSO回归是一种变量降维手段,其作用是从线性模型中筛选出对结果影响大的变量们。
可控制回归的参数主要是λ,λ不同得到的最终变量数量也不同。
图1主要是展示λ与变量数量的关系
图2主要是展示λ与似然偏差的关系,常选择λ[min-se]

(K折交叉验证的K也对结果有影响,但我还不太懂其意义)

原理

Attachments/Pasted image 20250926134824.png
Attachments/Pasted image 20250926134830.png
Attachments/Pasted image 20250926134907.png
Attachments/Pasted image 20250926140507.pngAttachments/Pasted image 20250926140647.pngAttachments/Pasted image 20250926140648.pngAttachments/Pasted image 20250926140756.pngAttachments/Pasted image 20250926212415.pngAttachments/Pasted image 20250926205907.png
Attachments/Pasted image 20250926210143.pngAttachments/Pasted image 20250926212356.png

应用

肿瘤:预后
非肿瘤:鉴别诊断

模型(prognostic nodel)/ signature/ riskscore

非肿瘤应用代码

肿瘤应用代码

建模/SCORE

Attachments/Pasted image 20250925085732.png

从网盘中或B站再找相关视频及代码