Create app.py

app.py

@@ -0,0 +1,60 @@
+import gradio as gr
+import pandas as pd
+import pickle
+import xgboost as xgb
+
+# 1. LOAD THE PICKLE FILE
+# Ensure 'model.pkl' is uploaded to your Hugging Face Space files
+with open("model.pkl", "rb") as f:
+    model = pickle.load(f)
+
+# 2. FULL FEATURE LIST (Ordered exactly as per your screenshot)
+ALL_FEATURES = [
+    'InternetService_1', 'Contract_0', 'tenure', 'InternetService_0',
+    'Contract_1', 'MultipleLines_1', 'PaperlessBilling_0', 'StreamingMovies_1',
+    'SeniorCitizen', 'PaymentMethod_0', 'TotalCharges', 'StreamingTV_1',
+    'MonthlyCharges', 'PaymentMethod_1', 'OnlineSecurity_1', 'TechSupport_1',
+    'MultipleLines_0', 'PhoneService_0', 'OnlineBackup_0', 'OnlineSecurity_0',
+    'StreamingMovies_0', 'StreamingTV_0', 'DeviceProtection_0', 
+    'OnlineBackup_1', 'DeviceProtection_1', 'TechSupport_0'
+]
+
+def predict_churn(tenure, internet_service_fiber, contract_month_to_month):
+    # Initialize all 26 features to 0
+    input_dict = {feature: [0.0] for feature in ALL_FEATURES}
+    
+    # Update the 3 features the user interacts with
+    input_dict['tenure'] = [float(tenure)]
+    input_dict['InternetService_1'] = [1.0 if internet_service_fiber else 0.0]
+    input_dict['Contract_0'] = [1.0 if contract_month_to_month else 0.0]
+    
+    # IMPORTANT: Set sensible defaults for key continuous variables 
+    # if they aren't provided by the user (prevents skewed results)
+    input_dict['MonthlyCharges'] = [65.0]
+    input_dict['TotalCharges'] = [2000.0]
+    
+    # Create DataFrame and ensure column order matches ALL_FEATURES exactly
+    input_data = pd.DataFrame(input_dict)[ALL_FEATURES]
+    
+    # 3. RUN PREDICTION
+    # We use predict_proba to get the confidence level
+    prediction_proba = model.predict_proba(input_data)[0][1]
+    prediction = "Churn Risk" if prediction_proba > 0.5 else "Stay"
+    
+    return f"Result: {prediction} (Confidence: {prediction_proba:.2%})"
+
+# 4. DEFINE THE UI
+demo = gr.Interface(
+    fn=predict_churn,
+    inputs=[
+        gr.Slider(0, 72, label="Tenure (Months)", value=12),
+        gr.Checkbox(label="Internet Service: Fiber Optic?"),
+        gr.Checkbox(label="Contract: Month-to-Month?")
+    ],
+    outputs="text",
+    title="Telco Churn Prediction Agent",
+    description="Using the model's top drivers to assess customer loyalty."
+)
+
+if __name__ == "__main__":
+    demo.launch()