http://blog.itpub.net/29829936/viewspace-2600454/
XAI是Github上的一个机器学习可解释性工具箱。XAI包含多种分析和评价数据和模型的工具。XAI在开发时遵循负责的机器学习的8个原则。
XAI是Github上的一个机器学习可解释性工具箱,地址为:
https://github.com/EthicalML/xai
安装及一些简单的用例如下:
安装
pip install xai简单用例
XAI可以识别数据不平衡。我们先载入census数据集:
import xai.datadf = xai.data.load_census()df.head()
查看多列类别不平衡:
protected_cols = ["gender", "ethnicity", "age"]ims = xai.show_imbalances(df, protected_cols)
查看一列类别不平衡:
im = xai.show_imbalance(df, "gender")
查看一列与另一列相交的不平衡:
im = xai.show_imbalance(df, "gender", cross=["loan"])
利用上采样或下采样进行平衡:
bal_df = xai.balance(df, "gender", cross=["loan"], upsample=1.0)
创建一个平衡的测试-训练划分:
# Balanced train-test split with minimum 300 examples of # the cross of the target y and the column genderx_train, y_train, x_test, y_test = xai.balanced_train_test_split( x, y, cross=["gender"], categorical_cols=categorical_cols, min_per_class=300)# Visualise the imbalances of gender and the target df_test = x_test.copy()df_test["loan"] = y_test_= xai.show_imbalance(df_test, "gender", cross=["loan"], categorical_cols=categorical_cols)
更多用例可以参考Github项目链接:
https://github.com/EthicalML/xai