Update app.py

app.py

@@ -1,59 +1,111 @@
 import pickle
 import pandas as pd
-import gradio as gr
 import shap
+from shap.plots._force_matplotlib import draw_additive_plot
+import gradio as gr
+import numpy as np
 import matplotlib.pyplot as plt
 
-# Load model and SHAP explainer
-model = pickle.load(open("default_xgb.pkl", "rb"))
-explainer = shap.Explainer(model)
-
-# Predict function using threshold = 0.4 and SHAP bar chart
-def predict_default(education, limit_bal, pay_0, pay_2, pay_3, pay_4, pay_5, pay_6):
-    input_data = pd.DataFrame([{
-        'EDUCATION': education,
-        'LIMIT_BAL': limit_bal,
-        'PAY_0': pay_0,
-        'PAY_2': pay_2,
-        'PAY_3': pay_3,
-        'PAY_4': pay_4,
-        'PAY_5': pay_5,
-        'PAY_6': pay_6
-    }])
-
-    # Predict probability and apply threshold
-    prob = model.predict_proba(input_data)[0, 1]
-    prediction = "Default" if prob >= 0.4 else "No Default"
-
-    # SHAP explanation
-    shap_values = explainer(input_data)
+# Load model
+loaded_model = pickle.load(open("default_xgb.pkl", 'rb'))
+
+# SHAP Explainer
+explainer = shap.Explainer(loaded_model)  # DO NOT CHANGE
+
+# Define main function
+def main_func(LIMIT_BAL, EDUCATION,
+              PAY_0, PAY_2, PAY_3, PAY_4, PAY_5, PAY_6,
+              b1, b2, b3, b4, b5, b6):
+
+    new_row = pd.DataFrame.from_dict({
+        'LIMIT_BAL': [LIMIT_BAL],
+        'EDUCATION': [EDUCATION],
+        'PAY_0': [PAY_0],
+        'PAY_2': [PAY_2],
+        'PAY_3': [PAY_3],
+        'PAY_4': [PAY_4],
+        'PAY_5': [PAY_5],
+        'PAY_6': [PAY_6],
+        'b1': [b1],
+        'b2': [b2],
+        'b3': [b3],
+        'b4': [b4],
+        'b5': [b5],
+        'b6': [b6]
+    })
+
+    prob = loaded_model.predict_proba(new_row)[0][1]
+    prediction = int(prob > 0.4)
+
+    shap_values = explainer(new_row)
     plt.figure()
-    shap.plots.bar(shap_values[0], max_display=7, show=False)
-    plt.tight_layout()
-    fig = plt.gcf()
+    shap.plots.bar(shap_values[0], max_display=10, show=False)
+    local_plot = plt.gcf()
     plt.close()
 
-    return f"Prediction: {prediction} (Probability: {prob:.2f})", fig
-
-# Gradio Interface
-demo = gr.Interface(
-    fn=predict_default,
-    inputs=[
-        gr.Dropdown([1, 2, 3, 4], label="Education Level"),
-        gr.Number(label="Credit Limit"),
-        gr.Number(label="Repayment Status Sept (PAY_0)"),
-        gr.Number(label="Repayment Status Aug (PAY_2)"),
-        gr.Number(label="Repayment Status July (PAY_3)"),
-        gr.Number(label="Repayment Status June (PAY_4)"),
-        gr.Number(label="Repayment Status May (PAY_5)"),
-        gr.Number(label="Repayment Status April (PAY_6)")
-    ],
-    outputs=[
-        gr.Text(label="Prediction Result"),
-        gr.Plot(label="SHAP Explanation")
-    ],
-    title="📊 Credit Default Predictor with SHAP",
-    description="An XGBoost model using Optuna hyperparameters and a 0.4 threshold to predict credit default risk. SHAP plot shows which features drive the prediction."
-)
+    return {"No Default": 1 - float(prob), "Default": float(prob)}, local_plot
+
+# App UI metadata
+title = "**Default Risk Predictor & Interpreter** 📊"
+description1 = """
+This app uses financial data such as credit limits, repayment behavior, and bill balances to estimate the probability that a person may **default on their next credit payment**.  
+It also includes an interactive SHAP plot that explains which features most influenced the prediction.
+"""
+description2 = """
+To use the app, either adjust the values or click one of the examples. Then click on **Analyze** to get predictions and interpretations. 🔍
+"""
+
+# Launch Gradio app
+with gr.Blocks(title=title) as demo:
+    gr.Markdown(f"## {title}")
+    gr.Markdown(description1)
+    gr.Markdown("---")
+    gr.Markdown(description2)
+    gr.Markdown("---")
+    
+    with gr.Row():
+        with gr.Column():
+            LIMIT_BAL = gr.Slider(label="Credit Limit (LIMIT_BAL)", minimum=10000, maximum=1000000, value=150000, step=5000)
+            EDUCATION = gr.Slider(label="Education Level (EDUCATION)", minimum=0, maximum=6, value=2, step=1)
+            PAY_0 = gr.Slider(label="PAY_0 (Most recent payment status)", minimum=-2, maximum=8, value=0, step=1)
+            PAY_2 = gr.Slider(label="PAY_2", minimum=-2, maximum=8, value=0, step=1)
+            PAY_3 = gr.Slider(label="PAY_3", minimum=-2, maximum=8, value=0, step=1)
+            PAY_4 = gr.Slider(label="PAY_4", minimum=-2, maximum=8, value=0, step=1)
+            PAY_5 = gr.Slider(label="PAY_5", minimum=-2, maximum=8, value=0, step=1)
+            PAY_6 = gr.Slider(label="PAY_6", minimum=-2, maximum=8, value=0, step=1)
+            b1 = gr.Slider(label="Balance 1 (b1)", minimum=-50000, maximum=50000, value=5000, step=500)
+            b2 = gr.Slider(label="Balance 2 (b2)", minimum=-50000, maximum=50000, value=4500, step=500)
+            b3 = gr.Slider(label="Balance 3 (b3)", minimum=-50000, maximum=50000, value=4700, step=500)
+            b4 = gr.Slider(label="Balance 4 (b4)", minimum=-50000, maximum=50000, value=4300, step=500)
+            b5 = gr.Slider(label="Balance 5 (b5)", minimum=-50000, maximum=50000, value=4000, step=500)
+            b6 = gr.Slider(label="Balance 6 (b6)", minimum=-50000, maximum=50000, value=3900, step=500)
+            
+            submit_btn = gr.Button("Analyze")
+
+        with gr.Column(visible=True, scale=1, min_width=600) as output_col:
+            label = gr.Label(label="Predicted Default Risk")
+            local_plot = gr.Plot(label="SHAP Explanation Plot")
+
+            submit_btn.click(
+                main_func,
+                [LIMIT_BAL, EDUCATION,
+                 PAY_0, PAY_2, PAY_3, PAY_4, PAY_5, PAY_6,
+                 b1, b2, b3, b4, b5, b6],
+                [label, local_plot],
+                api_name="Default_Risk"
+            )
+
+    gr.Markdown("### Example inputs to try:")
+    gr.Examples(
+        examples=[
+            [200000, 2, 0, 0, 0, 0, 0, 0, 5200, 5000, 4800, 4600, 4400, 4200],
+            [80000, 3, 2, 0, 2, 3, 2, 2, -3000, -2500, -2000, -1800, -1500, -1200]
+        ],
+        inputs=[LIMIT_BAL, EDUCATION, PAY_0, PAY_2, PAY_3, PAY_4, PAY_5, PAY_6,
+                b1, b2, b3, b4, b5, b6],
+        outputs=[label, local_plot],
+        fn=main_func,
+        cache_examples=True
+    )
 
 demo.launch()