add feature importance & decision plots

app.py

@@ -6,6 +6,7 @@ import matplotlib.pyplot as plt
 import numpy as np
 import shap
 from sklearn.pipeline import make_pipeline
+import seaborn as sns
 
 feature_names = ["Age", "BMI", "HbA1c", "Blood Glucose"]
 
@@ -67,9 +68,9 @@ with gr.Blocks() as demo:
                     summary_plot = gr.Plot()
             with gr.Row():
                 with gr.Column():
-                    force_plot = gr.Plot()
+                    importance_plot = gr.Plot()
                 with gr.Column():
-                    interaction_plot = gr.Plot()
+                    decision_plot = gr.Plot()
 
     def submit_inputs(
         age_input,
@@ -171,25 +172,18 @@ with gr.Blocks() as demo:
         )
 
         fig4 = plt.figure(figsize=(3, 3))
-        plt.title("SHAP Force Plot")
-        shap.initjs()
-        shap.force_plot(
+        feature_importances = rf_classifier.feature_importances_
+        plt.title("Feature Importances")
+        sns.barplot(x=feature_importances, y=["age", "bmi", "hba1c", "glucose"])
+
+        fig5 = plt.figure(figsize=(3, 3))
+        plt.title("SHAP Interaction Plot")
+        shap.decision_plot(
             explainer.expected_value[predicted_class],
             shap_values_for_predicted_class,
             df.iloc[0],
-            feature_names=["age", "bmi", "hba1c", "glucose"],
-            matplotlib=True,
         )
 
-        # fig4 = plt.figure(figsize=(3, 3))
-        # plt.title("SHAP Interaction Plot")
-        # shap_interaction_values = explainer.shap_interaction_values(transformed_df)
-        # shap.summary_plot(
-        #     shap_interaction_values,
-        #     features=transformed_df,
-        #     feature_names=["age", "bmi", "hba1c", "glucose"],
-        # )
-
         ## save user's data in hopsworks
         if consent_input == True:
             user_data_fg = fs.get_or_create_feature_group(
@@ -202,7 +196,7 @@ with gr.Blocks() as demo:
             user_data_df["diabetes"] = existent_info_input
             user_data_fg.insert(user_data_df)
             print("inserted new user data to hopsworks", user_data_df)
-        return res, fig, fig2, fig3, fig4
+        return res, fig, fig2, fig3, fig4, fig5
 
     btn.click(
         submit_inputs,
@@ -214,7 +208,7 @@ with gr.Blocks() as demo:
             existent_info_input,
             consent_input,
         ],
-        outputs=[output, mean_plot, waterfall_plot, summary_plot, force_plot],
+        outputs=[output, mean_plot, waterfall_plot, importance_plot, decision_plot],
     )
 
 demo.launch()

requirements.txt

@@ -5,4 +5,5 @@ pandas
 scikit-learn==1.1.1
 matplotlib
 numpy
-shap
+shap
+seaborn