Update app.py

app.py

@@ -1,176 +1,73 @@
-#  Gradio App 
-#Imports
 import gradio as gr
-import joblib
 import pandas as pd
-import numpy as np
+import joblib
 import shap
-import plotly.graph_objects as go
-from sklearn.preprocessing import StandardScaler
-# We'll ask user for human-friendly original features and then transform into the encoded features
-# Input fields based on original Telco columns (commonly present)
-def preprocess_user_input(user_inputs):
-    """
-    user_inputs: dict with original columns (human-friendly)
-    returns: DataFrame encoded to match feature_cols order
-    """
-    # Build single-row DataFrame with original categorical values
-    row = pd.DataFrame([user_inputs])
-    # same preprocessing as training: get_dummies with drop_first
-    row_enc = pd.get_dummies(row, drop_first=True)
-    # reindex to feature_cols with 0 fill for missing dummies
-    row_enc = row_enc.reindex(columns=feature_cols, fill_value=0)
-    # scale numeric columns
-    if numeric_cols:
-        row_enc[numeric_cols] = scaler.transform(row_enc[numeric_cols])
-    return row_enc
-
-def create_gauge(prob):
-    # prob is in [0,1]; create a Plotly gauge
-    fig = go.Figure(go.Indicator(
-        mode="gauge+number",
-        value=prob * 100,
-        number={'suffix': "%"},
-        title={'text': "Churn Probability"},
-        gauge={
-            'axis': {'range': [0, 100]},
-            'bar': {'color': "red" if prob > 0.6 else ("orange" if prob > 0.3 else "green")},
-            'steps': [
-                {'range': [0, 30], 'color': "lightgreen"},
-                {'range': [30, 60], 'color': "yellow"},
-                {'range': [60, 100], 'color': "lightcoral"}
-            ],
-        }
-    ))
-    fig.update_layout(height=300, margin=dict(l=20, r=20, t=40, b=20))
-    return fig
-
-def top_shap_barchart(shap_vals, feat_names, top_n=5):
-    # shap_vals: 1D array of shap values for the single prediction
-    abs_vals = np.abs(shap_vals)
-    idx = np.argsort(abs_vals)[::-1][:top_n]
-    top_feats = [feat_names[i] for i in idx]
-    top_vals = shap_vals[idx]
-    # Plotly horizontal bar
-    fig = go.Figure(go.Bar(
-        x=top_vals[::-1],
-        y=[f for f in top_feats[::-1]],
-        orientation='h'
-    ))
-    fig.update_layout(title=f"Top {top_n} SHAP feature contributions (positive increases churn)", height=300, margin=dict(l=100))
-    return fig
+import numpy as np
 
-# The actual predict function wired to Gradio
-def predict_and_explain(
-    gender, SeniorCitizen, Partner, Dependents,
-    tenure, PhoneService, MultipleLines, InternetService,
-    OnlineSecurity, OnlineBackup, DeviceProtection, TechSupport,
-    StreamingTV, StreamingMovies, Contract, PaperlessBilling, PaymentMethod,
-    MonthlyCharges, TotalCharges,
-    example_choice=None
-):
-    # If example button used, ignore other inputs and load preset
-    if example_choice == "High churn risk example":
-        inputs = {
-            "gender": "Male", "SeniorCitizen": "No", "Partner":"No", "Dependents":"No",
-            "tenure": 4, "PhoneService":"Yes", "MultipleLines":"No", "InternetService":"Fiber optic",
-            "OnlineSecurity":"No","OnlineBackup":"No","DeviceProtection":"No","TechSupport":"No",
-            "StreamingTV":"Yes","StreamingMovies":"Yes","Contract":"Month-to-month","PaperlessBilling":"Yes",
-            "PaymentMethod":"Electronic check","MonthlyCharges":90.0,"TotalCharges":360.0
-        }
-    elif example_choice == "Low churn risk example":
-        inputs = {
-            "gender": "Female", "SeniorCitizen": "No", "Partner":"Yes", "Dependents":"Yes",
-            "tenure": 48, "PhoneService":"Yes", "MultipleLines":"Yes", "InternetService":"DSL",
-            "OnlineSecurity":"Yes","OnlineBackup":"Yes","DeviceProtection":"Yes","TechSupport":"Yes",
-            "StreamingTV":"No","StreamingMovies":"No","Contract":"Two year","PaperlessBilling":"No",
-            "PaymentMethod":"Mailed check","MonthlyCharges":40.0,"TotalCharges":1920.0
-        }
-    else:
-        inputs = {
-            "gender": gender, "SeniorCitizen": SeniorCitizen, "Partner": Partner, "Dependents": Dependents,
-            "tenure": tenure, "PhoneService": PhoneService, "MultipleLines": MultipleLines, "InternetService": InternetService,
-            "OnlineSecurity": OnlineSecurity, "OnlineBackup": OnlineBackup, "DeviceProtection": DeviceProtection, "TechSupport": TechSupport,
-            "StreamingTV": StreamingTV, "StreamingMovies": StreamingMovies, "Contract": Contract, "PaperlessBilling": PaperlessBilling,
-            "PaymentMethod": PaymentMethod, "MonthlyCharges": MonthlyCharges, "TotalCharges": TotalCharges
-        }
+# Load model, scaler, and feature list
+model = joblib.load("churn_model.pkl")
+scaler = joblib.load("scaler.pkl")
+feature_cols = joblib.load("feature_cols.pkl")  # list of feature names
 
-    # Preprocess -> encoded row matching training features
-    row_enc = preprocess_user_input(inputs)
+# If you saved an explainer, load it; otherwise create new one
+try:
+    explainer = joblib.load("explainer.pkl")
+except:
+    explainer = shap.Explainer(model, feature_names=feature_cols)
 
-    # Predict
-    prob = float(model.predict_proba(row_enc)[0, 1])
-    pred_label = "⚠️ High Risk of Churn" if prob > 0.5 else "✅ Low Risk (likely to stay)"
 
-    # SHAP explanation
-    shap_vals = explainer.shap_values(row_enc)  # might return array shape (1, n_features) or list
-    # normalize into array
-    if isinstance(shap_vals, list):
-        shap_arr = np.array(shap_vals)[0]
-    else:
-        shap_arr = np.array(shap_vals).reshape(-1)
-    # Prepare top contributors text
-    abs_idx = np.argsort(np.abs(shap_arr))[::-1][:3]
-    top_contribs = []
-    for i in abs_idx:
-        sign = "+" if shap_arr[i] > 0 else "-"
-        top_contribs.append(f"{feature_cols[i]}: {sign}{abs(shap_arr[i]):.3f}")
-    top_text = "\n".join(top_contribs)
+# Preprocessing Function
+def preprocess_user_input(user_inputs):
+    """
+    user_inputs: list of values in correct order
+    returns: DataFrame ready for prediction
+    """
+    row = pd.DataFrame([user_inputs], columns=feature_cols)
+    row_scaled = scaler.transform(row)
+    return pd.DataFrame(row_scaled, columns=feature_cols)
 
-    # Create visuals
-    gauge_fig = create_gauge(prob)
-    shap_fig = top_shap_barchart(shap_arr, feature_cols, top_n=5)
 
-    # Business insight text
-    insight = f"{pred_label}\nChurn Probability: {prob:.2%}\n\nTop factors:\n{top_text}"
+# Prediction + SHAP Explanation
+def predict_and_explain(*inputs):
+    # Preprocess
+    row_scaled = preprocess_user_input(inputs)
+    
+    # Predict probability and class
+    pred_prob = model.predict_proba(row_scaled)[0][1]
+    pred_class = model.predict(row_scaled)[0]
+    
+    # SHAP values
+    shap_values = explainer(row_scaled)
+    
+    # Explanation as bar chart
+    shap_fig = shap.plots.bar(shap_values[0], show=False)
+    
+    result_text = f"**Prediction:** {'Churn' if pred_class==1 else 'No Churn'}\n"
+    result_text += f"**Probability of Churn:** {pred_prob:.2%}"
+    
+    return result_text, shap_fig
 
-    return insight, gauge_fig, shap_fig
 
-# UI layout with Gradio Blocks
+# Build Gradio UI
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# 📊 Customer Churn Prediction — Explainable Demo")
-    gr.Markdown("Fill in customer info or choose an example. The app returns a churn probability (gauge) and the top feature contributions (SHAP).")
-
+    gr.Markdown("## 📊 Customer Churn Prediction with SHAP Explanations")
+    
     with gr.Row():
-        with gr.Column(scale=1):
-            gender = gr.Dropdown(["Female","Male"], value="Male", label="Gender")
-            SeniorCitizen = gr.Dropdown(["No","Yes"], value="No", label="Senior Citizen")
-            Partner = gr.Dropdown(["No","Yes"], value="No", label="Partner")
-            Dependents = gr.Dropdown(["No","Yes"], value="No", label="Dependents")
-            tenure = gr.Slider(minimum=0, maximum=72, step=1, value=12, label="Tenure (months)")
-            MonthlyCharges = gr.Slider(minimum=0, maximum=200, step=0.5, value=70, label="Monthly Charges (USD)")
-            TotalCharges = gr.Number(value=700, label="Total Charges (USD)")
-            PhoneService = gr.Dropdown(["No","Yes"], value="Yes", label="Phone Service")
-            MultipleLines = gr.Dropdown(["No","Yes","No phone service"], value="No", label="Multiple Lines")
-            InternetService = gr.Dropdown(["DSL","Fiber optic","No"], value="Fiber optic", label="Internet Service")
-            OnlineSecurity = gr.Dropdown(["No","Yes","No internet service"], value="No", label="Online Security")
-            OnlineBackup = gr.Dropdown(["No","Yes","No internet service"], value="No", label="Online Backup")
-            DeviceProtection = gr.Dropdown(["No","Yes","No internet service"], value="No", label="Device Protection")
-            TechSupport = gr.Dropdown(["No","Yes","No internet service"], value="No", label="Tech Support")
-            StreamingTV = gr.Dropdown(["No","Yes","No internet service"], value="No", label="Streaming TV")
-            StreamingMovies = gr.Dropdown(["No","Yes","No internet service"], value="No", label="Streaming Movies")
-            Contract = gr.Dropdown(["Month-to-month","One year","Two year"], value="Month-to-month", label="Contract")
-            PaperlessBilling = gr.Dropdown(["No","Yes"], value="Yes", label="Paperless Billing")
-            PaymentMethod = gr.Dropdown(["Electronic check","Mailed check","Bank transfer (automatic)","Credit card (automatic)"],
-                                      value="Electronic check", label="Payment Method")
-            example_choice = gr.Radio(["None","High churn risk example","Low churn risk example"], value="None", label="Load example?")
-            submit_btn = gr.Button("Predict & Explain", variant="primary")
-
-        with gr.Column(scale=1):
-            out_text = gr.Textbox(label="Prediction & Business Insight", lines=7)
-            out_gauge = gr.Plot(label="Churn Probability Gauge")
-            out_shap = gr.Plot(label="Top SHAP Contributors")
-
-    # Hook up the button
+        inputs = []
+        for col in feature_cols:
+            inputs.append(gr.Number(label=col))
+    
+    output_text = gr.Markdown()
+    output_plot = gr.Plot()
+    
+    submit_btn = gr.Button("Predict")
+    
     submit_btn.click(
-        fn=predict_and_explain,
-        inputs=[gender, SeniorCitizen, Partner, Dependents,
-                tenure, PhoneService, MultipleLines, InternetService,
-                OnlineSecurity, OnlineBackup, DeviceProtection, TechSupport,
-                StreamingTV, StreamingMovies, Contract, PaperlessBilling, PaymentMethod,
-                MonthlyCharges, TotalCharges, example_choice],
-        outputs=[out_text, out_gauge, out_shap]
+        predict_and_explain,
+        inputs=inputs,
+        outputs=[output_text, output_plot]
     )
 
-# Launch demo (share=True gives public link)
-demo.launch(share=True)
+# Launch
+if __name__ == "__main__":
+    demo.launch()