6个机器学习可解释性框架!
Datawhale
共 2317字,需浏览 5分钟
·
2022-08-26 18:17
Datawhale干货
作者:Moez Ali,来源:DeepHub IMBA
如何学习的? 学到了什么? 针对一个特定输入为什么会做出如此决策? 决策是否可靠?
SHAP
基于随机森林模型的心脏病患者预测分类
![](https://filescdn.proginn.com/67341332f1e17739d76d29c7e980939d/8a847beb5a74763258a7ef7b0e76beaa.webp)
# install with pip
pip install shap
# install with conda
conda install -c conda-forge shap
![](https://filescdn.proginn.com/6c37d623e03b7a6d0fd2a6c7c21ebdf3/e753ae0950413e2c3c85010b9db1ba74.webp)
![](https://filescdn.proginn.com/941d2507e9dc7916434ac051c2537c09/d6e0c154bc998ea32be2b9bdf5b0f4c2.webp)
![](https://filescdn.proginn.com/ab5859edf210072f470237f8e350a4e0/c588cccd5139dbef7e2eddfba79a9626.webp)
LIME
![](https://filescdn.proginn.com/c9d44650ddd1985eefa29e60d1803427/02a2e09416120a4f27a1ffe3170c8707.webp)
pip install lime
![](https://filescdn.proginn.com/4dead343932f3d10439f0684584de7ff/822009a053855d00ff2610bc076f225e.webp)
Shapash
![](https://filescdn.proginn.com/924cdab1e363e9dda48dccee8e47ed9d/c5352b0dff5fb59e3f3f6e510bd907e9.webp)
![](https://filescdn.proginn.com/c2416d9bec52747fd000c590d656181c/0a9d04fa206b9f678fc67c539851701f.webp)
![](https://filescdn.proginn.com/7e6d34655025b9a1fd930f2824a16243/a48ecdf34c2264700e27bf47d7228dcb.webp)
InterpretML
![](https://filescdn.proginn.com/f6e84f9c00afadd18d81c300fc40cb20/c1100f4f3ba9c3aa86d7a4dd8e6621c6.webp)
![](https://filescdn.proginn.com/97e458667ba6b085ecd8e21a4d9e717f/cbedc00b486d4b0efe1ff47a898a3cd6.webp)
ELI5
scikit-learn XGBoost、LightGBM CatBoost Keras
检查模型参数并说明模型是如何全局工作的; 检查模型的单个预测并说明什么模型会做出这样的决定。
![](https://filescdn.proginn.com/579b016bef7b8cd92b1d380e27411ace/793d1e89eafa91892648a1f128bf1f72.webp)
![](https://filescdn.proginn.com/a20185e82b474debd181eff015563f16/8d1cf6d2cd74773f900e68d560c5b8a1.webp)
OmniXAI
![](https://filescdn.proginn.com/7ac73f8ef59e3e57a46a2de05709eb5b/8fd6adada7ecfaaa0153b856b67678e9.webp)
![](https://filescdn.proginn.com/6bf39e078aa750747e3d856c360428ac/9d5023c2b018732c811697f512b6f3f1.webp)
评论