Update app.py

app.py

@@ -1,15 +1,130 @@
 import gradio as gr
 import pickle
 import numpy as np
+import pandas as pd
 import matplotlib.pyplot as plt
+import matplotlib
+matplotlib.use("Agg")
 
-# Load model
-with open("DecisionTreeClassifier.pkl", "rb") as file:
-    model = pickle.load(file)
+# =========================
+# 🔹 Load Model
+# =========================
+try:
+    with open("DecisionTreeClassifier.pkl", "rb") as f:
+        model = pickle.load(f)
+except:
+    model = None
 
+# =========================
+# 🔹 Load Dataset (REAL or DUMMY)
+# =========================
+try:
+    data = pd.read_csv("churn_data.csv")  # 🔥 Replace with your dataset
+except:
+    # fallback dummy dataset
+    np.random.seed(42)
+    data = pd.DataFrame({
+        "gender": np.random.choice(["Male","Female"], 300),
+        "tenure": np.random.randint(1, 60, 300),
+        "MonthlyCharges": np.random.randint(500, 8000, 300),
+        "Contract": np.random.choice(["Monthly","Quarterly","Yearly"], 300),
+        "Churn": np.random.choice([0,1], 300)
+    })
+
+# =========================
+# 🔹 KPI FUNCTION
+# =========================
+def get_kpis(df):
+    total = len(df)
+    churn_rate = df["Churn"].mean() * 100
+    avg_spend = df["MonthlyCharges"].mean()
+    avg_tenure = df["tenure"].mean()
+
+    return f"""
+### 📊 Key Metrics
+- Total Customers: **{total}**
+- Churn Rate: **{churn_rate:.2f}%**
+- Avg Spend: **₹{avg_spend:.0f}**
+- Avg Tenure: **{avg_tenure:.1f} months**
+"""
+
+# =========================
+# 🔹 FILTER FUNCTION
+# =========================
+def apply_filters(gender, contract):
+    df = data.copy()
+
+    if gender != "All":
+        df = df[df["gender"] == gender]
+
+    if contract != "All":
+        df = df[df["Contract"] == contract]
+
+    return df
+
+# =========================
+# 🔹 CHARTS
+# =========================
+def churn_dist(df):
+    fig, ax = plt.subplots()
+    counts = df["Churn"].value_counts()
+    ax.bar(["No Churn","Churn"], counts)
+    ax.set_title("Churn Distribution")
+    plt.close(fig)
+    return fig
+
+def contract_chart(df):
+    fig, ax = plt.subplots()
+    pd.crosstab(df["Contract"], df["Churn"]).plot(kind="bar", ax=ax)
+    ax.set_title("Churn by Contract")
+    plt.close(fig)
+    return fig
+
+def tenure_chart(df):
+    fig, ax = plt.subplots()
+    ax.scatter(df["tenure"], df["Churn"])
+    ax.set_title("Tenure vs Churn")
+    plt.close(fig)
+    return fig
+
+def risk_pie(df):
+    fig, ax = plt.subplots()
+    counts = df["Churn"].value_counts()
+    ax.pie(counts, labels=["No Churn","Churn"], autopct="%1.1f%%")
+    ax.set_title("Risk Segmentation")
+    plt.close(fig)
+    return fig
+
+# =========================
+# 🔹 FEATURE IMPORTANCE
+# =========================
+def feature_importance():
+    if model is None or not hasattr(model, "feature_importances_"):
+        return None
+
+    features = [
+        "age","gender","tenure","usage","support","delay",
+        "spend","interaction",
+        "sub_premium","sub_standard",
+        "contract_monthly","contract_quarterly"
+    ]
+
+    fig, ax = plt.subplots()
+    ax.barh(features, model.feature_importances_)
+    ax.set_title("Feature Importance")
+    plt.close(fig)
+    return fig
+
+# =========================
+# 🔹 PREDICTION
+# =========================
 def predict_churn(age, gender, tenure, usage, support, delay,
-                  subscription, contract, spend, interaction):
+                 subscription, contract, spend, interaction):
+
     try:
+        if model is None:
+            return "Model not loaded ❌", "", "", None, ""
+
         gender_val = 1 if gender == "Female" else 0
 
         sub_premium = 1 if subscription == "Premium" else 0
@@ -26,11 +141,14 @@ def predict_churn(age, gender, tenure, usage, support, delay,
         ]])
 
         pred = model.predict(input_data)[0]
-        prob = model.predict_proba(input_data)[0][1]
+
+        if hasattr(model, "predict_proba"):
+            prob = model.predict_proba(input_data)[0][1]
+        else:
+            prob = 0.5
 
         result = "⚠️ Likely to Churn" if pred == 1 else "✅ Stable Customer"
 
-        # Risk label
         if prob > 0.7:
             risk = "🔴 High Risk"
         elif prob > 0.4:
@@ -38,61 +156,110 @@ def predict_churn(age, gender, tenure, usage, support, delay,
         else:
             risk = "🟢 Low Risk"
 
-        # Styled graph
+        # probability chart
         fig, ax = plt.subplots()
-        ax.bar(["No Churn", "Churn"], [1 - prob, prob])
-        ax.set_title("Churn Probability Analysis")
-        ax.set_ylim(0, 1)
+        ax.bar(["No Churn","Churn"], [1-prob, prob])
+        ax.set_ylim(0,1)
+        ax.set_title("Prediction Probability")
+        plt.close(fig)
+
+        # explanation
+        reasons = []
+        if delay > 15: reasons.append("High payment delay")
+        if tenure < 6: reasons.append("Low tenure")
+        if support > 5: reasons.append("Too many support calls")
+
+        explanation = "\n".join(reasons) if reasons else "No strong risk indicators"
 
-        return result, f"{prob*100:.2f}%", risk, fig
+        return result, f"{prob*100:.2f}%", risk, fig, explanation
 
     except Exception as e:
-        return f"Error: {str(e)}", "", "", None
+        return f"Error: {str(e)}", "", "", None, ""
 
+# =========================
+# 🎨 UI
+# =========================
+with gr.Blocks() as demo:
 
-# 🎨 Custom CSS
-custom_css = """
-body {background-color: #0f172a; color: white;}
-.gradio-container {max-width: 900px; margin: auto;}
-h1 {text-align: center; color: #38bdf8;}
-"""
+    gr.Markdown("# 🚀 Customer Churn Analytics Dashboard")
+
+    # ---------------------
+    # 📊 DASHBOARD TAB
+    # ---------------------
+    with gr.Tab("📊 Dashboard"):
+
+        gender_filter = gr.Dropdown(["All","Male","Female"], value="All", label="Filter by Gender")
+        contract_filter = gr.Dropdown(["All","Monthly","Quarterly","Yearly"], value="All", label="Filter by Contract")
+
+        kpi_text = gr.Markdown()
+        chart1 = gr.Plot()
+        chart2 = gr.Plot()
+        chart3 = gr.Plot()
+        chart4 = gr.Plot()
+
+        def update_dashboard(g, c):
+            df = apply_filters(g, c)
+            return (
+                get_kpis(df),
+                churn_dist(df),
+                contract_chart(df),
+                tenure_chart(df),
+                risk_pie(df)
+            )
+
+        gender_filter.change(update_dashboard, [gender_filter, contract_filter],
+                             [kpi_text, chart1, chart2, chart3, chart4])
+
+        contract_filter.change(update_dashboard, [gender_filter, contract_filter],
+                               [kpi_text, chart1, chart2, chart3, chart4])
 
-# 🚀 UI with Blocks
-with gr.Blocks(css=custom_css) as demo:
+        # initial load
+        demo.load(update_dashboard, [gender_filter, contract_filter],
+                  [kpi_text, chart1, chart2, chart3, chart4])
 
-    gr.Markdown("# 📊 Customer Churn Prediction Dashboard")
-    gr.Markdown("### Analyze customer behavior and predict churn risk")
+    # ---------------------
+    # 🔍 PREDICTION TAB
+    # ---------------------
+    with gr.Tab("🔍 Prediction"):
 
-    with gr.Row():
-        with gr.Column():
-            age = gr.Slider(18, 80, label="Age")
-            gender = gr.Radio(["Male", "Female"], label="Gender")
-            tenure = gr.Slider(0, 60, label="Tenure (Months)")
-            usage = gr.Slider(0, 50, label="Usage Frequency")
+        with gr.Row():
+            age = gr.Slider(18,80,value=30)
+            gender = gr.Radio(["Male","Female"], value="Male")
+            tenure = gr.Slider(0,60,value=12)
+            usage = gr.Slider(0,50,value=10)
 
-        with gr.Column():
-            support = gr.Slider(0, 20, label="Support Calls")
-            delay = gr.Slider(0, 30, label="Payment Delay")
-            subscription = gr.Radio(["Basic", "Standard", "Premium"], label="Subscription Type")
-            contract = gr.Radio(["Monthly", "Quarterly", "Yearly"], label="Contract Length")
+        with gr.Row():
+            support = gr.Slider(0,20,value=2)
+            delay = gr.Slider(0,30,value=5)
+            subscription = gr.Radio(["Basic","Standard","Premium"], value="Basic")
+            contract = gr.Radio(["Monthly","Quarterly","Yearly"], value="Monthly")
 
-    spend = gr.Slider(0, 10000, label="Total Spend")
-    interaction = gr.Slider(0, 100, label="Last Interaction")
+        spend = gr.Slider(0,10000,value=2000)
+        interaction = gr.Slider(0,100,value=20)
 
-    predict_btn = gr.Button("🔍 Predict Churn", variant="primary")
+        btn = gr.Button("Predict")
 
-    gr.Markdown("## 📈 Results")
+        result = gr.Textbox(label="Prediction")
+        prob = gr.Textbox(label="Probability")
+        risk = gr.Textbox(label="Risk Level")
+        graph = gr.Plot()
+        explanation = gr.Textbox(label="Why this prediction?")
 
-    result = gr.Textbox(label="Prediction")
-    prob = gr.Textbox(label="Churn Probability")
-    risk = gr.Textbox(label="Risk Level")
-    graph = gr.Plot()
+        btn.click(
+            predict_churn,
+            inputs=[age, gender, tenure, usage, support, delay,
+                    subscription, contract, spend, interaction],
+            outputs=[result, prob, risk, graph, explanation]
+        )
 
-    predict_btn.click(
-        fn=predict_churn,
-        inputs=[age, gender, tenure, usage, support, delay,
-                subscription, contract, spend, interaction],
-        outputs=[result, prob, risk, graph]
-    )
+    # ---------------------
+    # 📈 INSIGHTS TAB
+    # ---------------------
+    with gr.Tab("📈 Insights"):
+        gr.Markdown("### Feature Importance")
+        gr.Plot(feature_importance())
 
-demo.launch()
+# =========================
+# 🚀 LAUNCH
+# =========================
+demo.launch(debug=True)