add ice pdp

pages/Anchors.py

@@ -86,7 +86,7 @@ def explain_example(anchors_threshold, example_idx):
         class_names,
         feature_names,
         X_train.values,
-        categorical_names, seed=42)
+        categorical_names)
     
     # Explain the selected sample
     exp = explainer.explain_instance(X_test.values[example_idx], global_model.predict, threshold=anchors_threshold)

pages/ICE_and_PDP.py

@@ -0,0 +1,68 @@
+# Import Libraries
+import matplotlib.pyplot as plt
+import streamlit as st
+import src.prompt_config as prompt_params
+# Models
+import xgboost
+from sklearn.model_selection import train_test_split
+from sklearn.inspection import PartialDependenceDisplay
+# XAI (Explainability)
+# import shap
+
+# Global Variables to Store Model & Data
+global_model = None
+X_train, X_test, y_train, y_test = None, None, None, None
+
+
+def train_model():
+    """ Train the XGBoost model only once and store it globally. """
+    global global_model, X_train, X_test, y_train, y_test
+    
+    if global_model is None:
+        # Load Data from SHAP library
+        X, y = shap.datasets.adult()
+        
+        # Split data
+        X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5, random_state=42)
+        
+        # Train XGBoost model
+        global_model = xgboost.XGBClassifier()
+        global_model.fit(X_train, y_train)
+        
+        print("XGBoost Model training completed!")
+
+def explain_example(features, kind):
+    """ Explain a given sample without retraining the model. """
+    global global_model, X_train, X_test, y_train, y_test
+    
+    if global_model is None:
+        train_model()
+    
+    fig, ax = plt.subplots(figsize=(10, 5))
+    PartialDependenceDisplay.from_estimator(global_model, X_test, features, kind=kind)
+    
+    st.pyplot(fig)
+
+def main():
+    global global_model
+    
+    # Ensure the model is trained only once
+    if global_model is None:
+        train_model()
+    # Define feature names
+    feature_names = ["Age", "Workclass", "Education-Num", "Marital Status", "Occupation",
+                     "Relationship", "Race", "Sex", "Capital Gain", "Capital Loss", "Hours per week", "Country"]
+
+    selected_feature = st.sidebar.selectbox("Select a feature for PDP/ICE analysis:", feature_names)
+    
+    kind = st.sidebar.selectbox("Select plot type:", ["average", "both", "individual"])
+    
+    st.title("ICE (Individual Conditional Expectation) and PDP (Partial Dependence Plot)")
+    st.write(prompt_params.ICE_INTRODUCTION)
+    # Explain the selected sample
+    if st.button("Explain Sample"):
+        explain_example(selected_feature, kind)
+
+
+if __name__ == '__main__':
+    main()

pages/SHAP.py

@@ -92,7 +92,7 @@ def main():
         label="Select the sample index to explain:",
         min_value=0,
         max_value=len(X_test) - 1,  # Ensures the index is within range
-        value=1,  # Default value
+        value=100,  # Default value
         step=1,  # Step size
         help=prompt_params.EXAMPLE_BE_EXPLAINED_IDX,
     )

src/prompt_config.py

@@ -77,4 +77,32 @@ The process of SHAP includes the following steps:
 4. **Ensure Additivity**: The sum of SHAP values should match the model's prediction difference from the baseline.
 
 By using SHAP, **interpretability** improves by generating stable and mathematically sound explanations, making models more transparent and trustworthy.
+"""
+
+ICE_INTRODUCTION = """
+Individual Conditional Expectation (ICE) plots provide a more granular view of feature influence by displaying the response of each individual instance to changes in a selected feature, rather than averaging across all instances as in Partial Dependence Plots (PDP).
+
+The process of ICE includes the following steps:
+1. **Select the Feature of Interest**: Choose a variable to analyze, such as Age.
+2. **Create a Feature Grid**: Define a range of values for the chosen feature (e.g., Age from 20 to 80).
+3. **Iterate Over Each Instance**: For each sample, replace the feature with values from the grid while keeping all other features unchanged.
+4. **Compute Predictions**: Use the trained model to predict outcomes for each modified sample and store the predictions.
+5. **Plot Individual Curves**: Each sample produces a separate curve, representing how its prediction evolves as the feature changes.
+
+By using ICE, **interpretability** improves by showing how a feature influences predictions at an individual level, capturing heterogeneous effects that PDP might average out.
+
+A Partial Dependence Plot (PDP) illustrates the marginal effect of a selected feature on the model’s predictions while averaging out the influence of all other features.
+
+The process of PDP includes the following steps:
+1. **Select the Feature of Interest**: Choose a variable for PDP analysis, such as Age.
+2. **Create a Feature Grid**: Define a range of values for the selected feature (e.g., Age from 20 to 80).
+3. **Modify the Dataset and Compute Predictions**: For each value in the grid, replace the feature value in all instances while keeping other features unchanged. Use the trained model to predict outcomes for the modified dataset.
+4. **Compute the Average Prediction**: Aggregate predictions across all instances and calculate the mean for each feature value in the grid.
+
+By using PDP, **interpretability** improves by showing the average effect of a feature on model predictions, making complex models more explainable.
+
+When `kind` is selected:
+- **both**: Displays both ICE and PDP.
+- **individual**: Displays only ICE.
+- **average**: Displays only PDP.
 """