如何在Matplotlib子图中并排绘制SHAP summary_plot？

SHAP的shap.summary_plot()默认会自动创建新的绘图窗口，且不支持直接传入ax参数指定子图，这导致你的代码会生成两个独立窗口。下面提供两种可行的解决方法：

方法1：使用SHAP的独立绘图函数（推荐，适用于SHAP 0.40+版本） #

SHAP较新版本提供了shap.plots.bar()和shap.plots.beeswarm()两个独立函数，它们支持传入ax参数来指定绘图的子图轴对象，完美适配Matplotlib的subplot布局。

修改后的代码如下：

import pandas as pd
import shap
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
from sklearn.preprocessing import LabelEncoder

mylabel = LabelEncoder()
data = pd.read_csv("https://raw.githubusercontent.com/krishnaik06/Multiple-Linear-Regression/master/50_Startups.csv")
data['State'] = mylabel.fit_transform(data['State'])

model = RandomForestRegressor()
y = data['Profit']
X = data.drop('Profit', axis=1)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=1)
model.fit(X_train, y_train)

explainer = shap.TreeExplainer(model)
shap_values = explainer.shap_values(X_train)

# 创建画布和子图
fig, (ax1, ax2) = plt.subplots(2, 1, figsize=(12, 15))

# 绘制条形图到ax1
shap.plots.bar(shap_values, feature_names=X.columns, ax=ax1)
# 绘制蜂群图到ax2
shap.plots.beeswarm(shap_values, feature_names=X.columns, ax=ax2)

plt.tight_layout()
plt.show()

方法2：切换当前活动子图（兼容旧版SHAP） #

如果你的SHAP版本较低，没有上述独立绘图函数，可以通过plt.sca()切换当前活动的子图轴，让shap.summary_plot()绘制到指定的子图中：

import pandas as pd
import shap
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
from sklearn.preprocessing import LabelEncoder

mylabel = LabelEncoder()
data = pd.read_csv("https://raw.githubusercontent.com/krishnaik06/Multiple-Linear-Regression/master/50_Startups.csv")
data['State'] = mylabel.fit_transform(data['State'])

model = RandomForestRegressor()
y = data['Profit']
X = data.drop('Profit', axis=1)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, random_state=1)
model.fit(X_train, y_train)

explainer = shap.TreeExplainer(model)
shap_values = explainer.shap_values(X_train)

# 创建画布和子图
plt.figure(figsize=(12, 15))

# 切换到第一个子图，绘制条形图
ax1 = plt.subplot(2, 1, 1)
plt.sca(ax1)
shap.summary_plot(shap_values, X_train, feature_names=X.columns, plot_type="bar", show=False)

# 切换到第二个子图，绘制蜂群图
ax2 = plt.subplot(2, 1, 2)
plt.sca(ax2)
shap.summary_plot(shap_values, X_train, feature_names=X.columns, show=False)

plt.tight_layout()
plt.show()

注意：此方法需要在调用shap.summary_plot()时加上show=False参数，避免提前显示绘图，最后统一用plt.show()展示整个画布。

内容的提问来源于stack exchange，提问作者user4356954