用LIME解释深层神经网络进行欺诈检测

Question

我建立了一个深度神经网络来分类欺诈性交易。我试图用LIME来解释，但是遇到了来自interpretor.explain_instance()函数的错误。

完整的代码如下：

import lime
from lime import lime_tabular

interpretor = lime_tabular.LimeTabularExplainer(
    training_data=x_train_scaled,
    feature_names=X_train.columns,
    mode='classification'
)

exp = interpretor.explain_instance(
    data_row=x_test_scaled[:1], ##new data
    predict_fn=model.predict,num_features=11
)
xp.show_in_notebook(show_table=True)

这会引发错误：

--
IndexError                                Traceback (most recent call last)
/tmp/ipykernel_33/1730959582.py in <module>
      1 exp = interpretor.explain_instance(
      2     data_row=x_test_scaled[1], ##new data
----> 3     predict_fn=model.predict
      4 )
      5 

/opt/conda/lib/python3.7/site-packages/lime/lime_tabular.py in explain_instance(self, data_row, predict_fn, labels, top_labels, num_features, num_samples, distance_metric, model_regressor)
    457                     num_features,
    458                     model_regressor=model_regressor,
--> 459                     feature_selection=self.feature_selection)
    460 
    461         if self.mode == "regression":

/opt/conda/lib/python3.7/site-packages/lime/lime_base.py in explain_instance_with_data(self, neighborhood_data, neighborhood_labels, distances, label, num_features, feature_selection, model_regressor)
    180 
    181         weights = self.kernel_fn(distances)
--> 182         labels_column = neighborhood_labels[:, label]
    183         used_features = self.feature_selection(neighborhood_data,
    184                                                labels_column,

IndexError: index 1 is out of bounds for axis 1 with size 1

Answer 1

在labels=(0,)中添加exp = eplainer.explain_instance()可以解决您的问题。

exp = interpretor.explain_instance(
data_row=x_test_scaled[:1], ##new data
predict_fn=model.predict,
num_features=11,
labels=(0,)
)

我也有一个类似的问题，乳腺癌数据试图预测良性肿瘤或恶性肿瘤。包含记录的示例的列标题为benign_0_malignant_1，每行中放置0或1。

Answer 2

我认为问题在于您传递的是一个2D数组，但是根据医生的说法 explain_instance()希望实例是一个一维数组。

请注意，来自2D数组的单行片本身就是2D：

>>> import numpy as np
>>> arr = np.array([[1, 2], [3, 4], [5, 6]])
>>> arr[:1]
array([[1, 2]])

另一个问题是预测函数，它将产生一系列概率，而不是单个预测。再一次，医生们解释(我的重点)：

classifier_fn -分类器预测概率函数，该函数采用numpy数组并输出预测概率.对于ScikitClassifiers，这是classifier.predict_proba。

要解决这些问题，请使用一个简单的索引，而不是切片，并将x_test_scaled作为预测函数传递给model.predict_proba：

exp = interpretor.explain_instance(data_row=x_test_scaled[0],
                                   predict_fn=model.predict_proba,
                                   num_features=11
                                   )