尝试从XgBoost显示SHAP值时出现predict（）问题？

Question

我运行了一个XgBoost模型，我想显示预测的SHAP值。我在为SHAP创建变量时遇到错误。这是我适合的XgBoost模型的代码：

reg = xgb.XGBRegressor(n_estimators=1000)
reg.fit(train_X, train_y,
    eval_set=[(train_X, train_y), (test_X, test_y)],
    early_stopping_rounds=50,
   verbose=False)

df_compare = pd.DataFrame(data=test_y, columns = ["actual"])
df_compare['predicted'] = reg.predict(test_X)

# Model Evaluation
sqrt(mean_squared_error(y_true=df_compare['actual'],
               y_pred=df_compare['predicted']))

# load JS visualization code to notebook
shap.initjs()

explainer = shap.TreeExplainer(reg)
shap_values = explainer.shap_values(test_X)

# summarize the effects of all the features
shap.summary_plot(shap_values, test_X)

shap_values变量上突出显示错误，错误代码为：

TypeError: predict() got an unexpected keyword argument 'validate_features'

我的目标是展示个体预测因子对测试集的贡献。这可以通过基于slundberg的GitHub存储库的“shap.summary_plot（）”命令来完成。

从我最初的研究来看，它似乎是XgBoost的一个常见问题，我想知道是否有人对此问题有任何解决方案？

任何帮助都会很棒！

编辑：以下是当前架构中test_X的示例：

array([[6.13181152e-01, 1.65250069e-01, 6.28375079e-01, 1.65250069e-01,
    7.69355058e-01, 1.65250069e-01, 4.00000000e+00, 1.20000000e+01,
    2.01300000e+03],
   [6.25013774e-01, 1.50569938e-01, 6.40500901e-01, 1.50569938e-01,
    7.84201386e-01, 1.50569938e-01, 1.00000000e+00, 1.00000000e+00,
    2.01400000e+03],
   [6.35163552e-01, 1.33475880e-01, 6.50902178e-01, 1.33475880e-01,
    7.96936256e-01, 1.33475880e-01, 1.00000000e+00, 2.00000000e+00,
    2.01400000e+03],
   [6.46226644e-01, 1.09757193e-01, 6.62239401e-01, 1.09757193e-01,
    8.10817057e-01, 1.09757193e-01, 1.00000000e+00, 3.00000000e+00,
    2.01400000e+03],
   [6.59526768e-01, 8.31406390e-02, 6.75869086e-01, 8.31406390e-02,
    8.27504651e-01, 8.31406390e-02, 2.00000000e+00, 4.00000000e+00,
    2.01400000e+03],
   [6.75320666e-01, 6.19388504e-02, 6.92054339e-01, 6.19388504e-02,
    8.47321169e-01, 6.19388504e-02, 2.00000000e+00, 5.00000000e+00,
    2.01400000e+03],
   [6.93341542e-01, 5.11984019e-02, 7.10521752e-01, 5.11984019e-02,
    8.69931864e-01, 5.11984019e-02, 2.00000000e+00, 6.00000000e+00,
    2.01400000e+03],
   [7.10885315e-01, 4.83581090e-02, 7.28500240e-01, 4.83581090e-02,
    8.91943941e-01, 4.83581090e-02, 3.00000000e+00, 7.00000000e+00,
    2.01400000e+03],
   [7.24623815e-01, 4.81424976e-02, 7.42579164e-01, 4.81424976e-02,
    9.09181562e-01, 4.81424976e-02, 3.00000000e+00, 8.00000000e+00,
    2.01400000e+03],
   [7.32223979e-01, 4.68193402e-02, 7.50367651e-01, 4.68193402e-02,
    9.18717446e-01, 4.68193402e-02, 3.00000000e+00, 9.00000000e+00,
    2.01400000e+03],
   [7.36887811e-01, 4.51536143e-02, 7.55147047e-01, 4.51536143e-02,
    9.24569131e-01, 4.51536143e-02, 4.00000000e+00, 1.00000000e+01,
    2.01400000e+03],
   [7.43107813e-01, 4.53410592e-02, 7.61521174e-01, 4.53410592e-02,
    9.32373334e-01, 4.53410592e-02, 4.00000000e+00, 1.10000000e+01,
    2.01400000e+03],
   [7.53861886e-01, 4.90621338e-02, 7.72541721e-01, 4.90621338e-02,
    9.45866411e-01, 4.90621338e-02, 4.00000000e+00, 1.20000000e+01,
    2.01400000e+03],
   [7.67586715e-01, 5.63629131e-02, 7.86606635e-01, 5.63629131e-02,
    9.63086879e-01, 5.63629131e-02, 1.00000000e+00, 1.00000000e+00,
    2.01500000e+03],
   [7.80160005e-01, 6.59919566e-02, 7.99491477e-01, 6.59919566e-02,
    9.78862518e-01, 6.59919566e-02, 1.00000000e+00, 2.00000000e+00,
    2.01500000e+03],
   [7.89674219e-01, 7.78638363e-02, 8.09241442e-01, 7.78638363e-02,
    9.90799950e-01, 7.78638363e-02, 1.00000000e+00, 3.00000000e+00,
    2.01500000e+03],
   [7.95533832e-01, 9.25097947e-02, 8.15246251e-01, 9.25097947e-02,
    9.98151976e-01, 9.25097947e-02, 2.00000000e+00, 4.00000000e+00,
    2.01500000e+03],
   [7.97006720e-01, 1.09847565e-01, 8.16755635e-01, 1.09847565e-01,
    1.00000000e+00, 1.09847565e-01, 2.00000000e+00, 5.00000000e+00,
    2.01500000e+03],
   [7.94528301e-01, 1.28832231e-01, 8.14215803e-01, 1.28832231e-01,
    9.96890340e-01, 1.28832231e-01, 2.00000000e+00, 6.00000000e+00,
    2.01500000e+03]])

Answer 1

它对我有用：

!pip install shap==0.19.2