SHAPs force_plot 不使用 DeepExplainer 绘图，但与 TreeExplainer 一起使用

如何解决SHAPs force_plot 不使用 DeepExplainer 绘图，但与 TreeExplainer 一起使用

虽然我已经能够让 SHAPs force_plot 与 TreeExplainer 一起使用，但我无法通过 DeepExplainer 显示它。我可以让它与 TreeExplainer 一起使用 matplotlib=True。但是，如果我尝试将 matplotlib 与 DeepExplainer 一起使用，则会出现错误：

“C:....\shap\plots_force_matplotlib.py”，第 236 行，格式数据 i[0] = convert_func(neg_val)

ValueError: 使用序列设置数组元素"

我在 PyCharm 中运行它。看起来 shap.initjs() 没有做任何事情，所以我尝试使用 matplotlib。我知道我应该尝试在 Jupyter 中为 initjs() 运行它，但我想看看我是否可以在 PyCharm 中运行它。

代码应该按原样在 Python 3+ 中运行，我还注释掉了 DeepExplainer 的 matplotlib 用法。

import numpy as np
np.random.seed(0)
from sklearn.model_selection import train_test_split
import shap
import matplotlib.pyplot as plt
import tensorflow as tf
import keras
from keras.layers import Dense
from sklearn.datasets import make_classification
from sklearn.ensemble import RandomForestRegressor

def build_model(num_features):
    model = keras.Sequential([
        Dense(16,activation=tf.nn.relu,input_shape=[num_features]),Dense(16,activation=tf.nn.relu),Dense(1)
    ])

    optimizer = tf.keras.optimizers.RMSprop()
    model.compile(loss='mse',optimizer=optimizer,metrics=['mae'])
    return model

class PrintDot(keras.callbacks.Callback):
  def on_epoch_end(self,epoch,logs):
    if epoch % 100 == 0: print('')
    print('.',end='')

number_of_features = 5
X,Y = make_classification(n_samples = 500,n_features = number_of_features,n_informative = 5,n_redundant = 0,n_classes = 3,weights = [.2,.3,.8])
X_train,X_test,Y_train,Y_test = train_test_split(X,Y,test_size = 0.3)

X_train_array,X_test_array,Y_train_array,Y_test_array =(np.asarray(X_train),np.asarray(X_test),np.asarray(Y_train),np.asarray(Y_test))

#SHAP displays force_plot for Random Forrests (plots).
model_RF = RandomForestRegressor(max_depth=6,random_state=0,n_estimators=10)
model_RF.fit(X_train,Y_train)
data_for_prediction = X_test_array[[0]]
explainer = shap.TreeExplainer(model_RF)
shap_values = explainer.shap_values(X_train_array)
p = shap.force_plot(explainer.expected_value[0],shap_values[0],data_for_prediction,matplotlib=True)


model_NN = build_model(number_of_features)
model_NN.summary()
early_stop = keras.callbacks.EarlyStopping(monitor='val_loss',patience=50)
history = model_NN.fit(X_train_array,epochs=10,validation_split = 0.2,verbose=0,callbacks=[early_stop,PrintDot()])
shap.initjs()

#SHAP force_plot (not displaying) for DeepExplainer - It runs throgh but doesn't display
data_for_prediction = X_test_array[[0]]
background = X_train_array
explainer = shap.DeepExplainer(model_NN,background)
shap_values = explainer.shap_values(data_for_prediction)
shap.force_plot(explainer.expected_value[0],data_for_prediction)
# Trying to use matplotlib=True as is done with Random Forrest plotting results in an error.
# shap.force_plot(explainer.expected_value[0],matplotlib=True)

# SHAP displays summary_plot
explainer = shap.DeepExplainer(model_NN,X_train_array)
shap_values = explainer.shap_values(X_test_array)
shap.summary_plot(shap_values[0],X_test_array)

感谢您的建议。