2024/03/09/16:00

analysis/others/shap_model.py

@@ -7,7 +7,7 @@ from classes.static_custom_class import StaticValue
 
 def draw_shap_beeswarm(model, x, feature_names, type, paint_object):
     plt.clf()
-    x = shap.sample(x, min(20, len(x)), random_state=StaticValue.RANDOM_STATE)
+    x = shap.sample(x, min(StaticValue.SAMPLE_NUM, len(x)), random_state=StaticValue.RANDOM_STATE)
     explainer = shap.KernelExplainer(model.predict, x)
     shap_values = explainer(x)
 
@@ -21,7 +21,7 @@ def draw_shap_beeswarm(model, x, feature_names, type, paint_object):
 
 def draw_waterfall(model, x, feature_names, number, paint_object):
     plt.clf()
-    x = shap.sample(x, min(20, len(x)), random_state=StaticValue.RANDOM_STATE)
+    x = shap.sample(x, min(StaticValue.SAMPLE_NUM, len(x)), random_state=StaticValue.RANDOM_STATE)
     explainer = shap.KernelExplainer(model.predict, x, feature_names=feature_names)
     shap_values = explainer(x)
 
@@ -35,7 +35,7 @@ def draw_waterfall(model, x, feature_names, number, paint_object):
 
 def draw_force(model, x, feature_names, number, paint_object):
     plt.clf()
-    x = shap.sample(x, min(20, len(x)), random_state=StaticValue.RANDOM_STATE)
+    x = shap.sample(x, min(StaticValue.SAMPLE_NUM, len(x)), random_state=StaticValue.RANDOM_STATE)
     explainer = shap.KernelExplainer(model.predict, x, feature_names=feature_names)
     shap_values = explainer(x[number])
 
@@ -49,7 +49,7 @@ def draw_force(model, x, feature_names, number, paint_object):
 
 def draw_dependence(model, x, feature_names, col, paint_object):
     plt.clf()
-    x = shap.sample(x, min(20, len(x)), random_state=StaticValue.RANDOM_STATE)
+    x = shap.sample(x, min(StaticValue.SAMPLE_NUM, len(x)), random_state=StaticValue.RANDOM_STATE)
     explainer = shap.KernelExplainer(model.predict, x, feature_names=feature_names)
     shap_values = explainer(x)
 

app.py

@@ -387,6 +387,9 @@ class Dataset:
 
         cls.container_dict = get_container_dict()
 
+        visualize = ""
+        choose_optimize = ""
+
     @classmethod
     def check_descriptive_indicators_df(cls):
         return True if not cls.descriptive_indicators_df.empty else False
@@ -1205,7 +1208,7 @@ class Dataset:
         EACH_ROW_NUM = 6 - 1
         output_list = []
 
-        if cls.cur_model and cls.choose_optimize:
+        if cls.cur_model and cls.check_model_optimize_radio():
             output_dict = ChooseModelParams.choose(cls.cur_model)
             row_unit_num_list = []
             row_len = len(output_dict.keys())
@@ -1531,12 +1534,12 @@ def get_return(is_visible, extra_gr_dict: dict = None):
             data_fit_button: gr.Button(LN.data_fit_button, visible=Dataset.check_before_train()),
             waterfall_radio: gr.Radio(Dataset.get_trained_model_list(), visible=Dataset.check_before_train(),
                                       label=LN.waterfall_radio),
-            waterfall_number: gr.Slider(0, 20, value=0, step=1,
+            waterfall_number: gr.Slider(0, StaticValue.SAMPLE_NUM, value=0, step=1,
                                         visible=Dataset.check_before_train(), label=LN.waterfall_number),
             waterfall_button: gr.Button(LN.waterfall_button, visible=Dataset.check_before_train()),
             force_radio: gr.Radio(Dataset.get_trained_model_list(), visible=Dataset.check_before_train(),
                                   label=LN.force_radio),
-            force_number: gr.Slider(0, 20, value=0, step=1,
+            force_number: gr.Slider(0, StaticValue.SAMPLE_NUM, value=0, step=1,
                                     visible=Dataset.check_before_train(), label=LN.force_number),
             force_button: gr.Button(LN.force_button, visible=Dataset.check_before_train()),
             dependence_radio: gr.Radio(Dataset.get_trained_model_list(), visible=Dataset.check_before_train(),

classes/static_custom_class.py

@@ -1,5 +1,7 @@
 # 全局静态变量值存储类
 class StaticValue:
+    # SHAP抽样数量
+    SAMPLE_NUM = 20
     # 超参数文本框的最大组件数量
     MAX_PARAMS_NUM = 60
     # 颜色和标签显示的最大组件数量