Upload 3 files

app.py

@@ -0,0 +1,92 @@
+import gradio as gr
+import pandas as pd
+import joblib
+import shap
+import matplotlib
+import matplotlib.pyplot as plt
+
+# Gunakan backend 'Agg' agar matplotlib aman dijalankan di server web
+matplotlib.use('Agg')
+
+# ==========================================
+# 1. LOAD MODEL & EXPLAINER
+# ==========================================
+print("Memuat model ChurnGuard...")
+model = joblib.load('churn_model.pkl')
+explainer = shap.Explainer(model)
+
+# ==========================================
+# 2. FUNGSI PREDIKSI & SHAP EXPLANATION
+# ==========================================
+def predict_and_explain(credit_score, age, tenure, balance, num_products, has_crcard, is_active, salary):
+    # 1. Format input user ke dalam DataFrame (Nama kolom WAJIB sama dengan saat training)
+    input_data = pd.DataFrame([[
+        credit_score, age, tenure, balance, num_products, has_crcard, is_active, salary
+    ]], columns=['CreditScore', 'Age', 'Tenure', 'Balance', 'NumOfProducts', 'HasCrCard', 'IsActiveMember', 'EstimatedSalary'])
+    
+    # 2. Lakukan Prediksi
+    prob_churn = model.predict_proba(input_data)[0][1]
+    prob_retain = 1 - prob_churn
+    
+    if prob_churn > 0.5:
+        keputusan = "### ⚠️ RISIKO TINGGI: Pelanggan diprediksi akan CHURN (Berhenti)."
+    else:
+        keputusan = "### ✅ AMAN: Pelanggan diprediksi RETAIN (Bertahan)."
+        
+    # 3. Generate Grafik SHAP (Explainable AI)
+    shap_values = explainer(input_data)
+    
+    plt.figure(figsize=(8, 4))
+    # Membuat waterfall plot untuk melihat fitur mana yang mendorong probabilitas churn
+    shap.plots.waterfall(shap_values[0], show=False)
+    fig = plt.gcf()
+    plt.tight_layout()
+    
+    return keputusan, {"Churn": prob_churn, "Retain": prob_retain}, fig
+
+# ==========================================
+# 3. ANTARMUKA GRADIO
+# ==========================================
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("""
+    <h1 style='text-align: center;'>🛡️ ChurnGuard: AI Customer Retention</h1>
+    <p style='text-align: center;'>Aplikasi Machine Learning (XGBoost) untuk memprediksi probabilitas pelanggan bank yang akan berhenti, lengkap dengan analisis <b>Explainable AI (SHAP)</b>.</p>
+    """)
+    
+    with gr.Row():
+        # BAGIAN KIRI: Form Input
+        with gr.Column(scale=1):
+            gr.Markdown("**👤 Profil Pelanggan**")
+            age = gr.Slider(18, 100, value=35, step=1, label="Umur (Age)")
+            tenure = gr.Slider(0, 10, value=3, step=1, label="Lama Menjadi Nasabah (Tenure - Tahun)")
+            balance = gr.Number(value=50000, label="Saldo Tabungan (Balance - $)")
+            salary = gr.Number(value=60000, label="Estimasi Gaji Tahunan (EstimatedSalary - $)")
+            
+            gr.Markdown("**📊 Status & Produk**")
+            credit_score = gr.Slider(300, 850, value=650, step=1, label="Skor Kredit (CreditScore)")
+            num_products = gr.Slider(1, 4, value=1, step=1, label="Jumlah Produk Bank yang Dipakai")
+            
+            # Ubah input Yes/No menjadi 1/0 agar sesuai dengan model
+            has_crcard = gr.Radio(choices=[1, 0], value=1, label="Punya Kartu Kredit? (1=Ya, 0=Tidak)")
+            is_active = gr.Radio(choices=[1, 0], value=1, label="Member Aktif? (1=Ya, 0=Tidak)")
+            
+            btn_predict = gr.Button("🔮 Prediksi & Analisis Risiko", variant="primary")
+            
+        # BAGIAN KANAN: Output Hasil & Analisis
+        with gr.Column(scale=1):
+            gr.Markdown("**📈 Hasil Keputusan AI**")
+            out_keputusan = gr.Markdown()
+            out_prob = gr.Label(label="Probabilitas")
+            
+            gr.Markdown("**🧠 Mengapa AI Memutuskan Demikian? (SHAP Analysis)**")
+            out_plot = gr.Plot(label="Analisis Fitur Pendukung Churn")
+            
+    # Hubungkan tombol dengan fungsi utama
+    btn_predict.click(
+        fn=predict_and_explain, 
+        inputs=[credit_score, age, tenure, balance, num_products, has_crcard, is_active, salary], 
+        outputs=[out_keputusan, out_prob, out_plot]
+    )
+
+if __name__ == "__main__":
+    demo.launch()

churn_model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7402d6e3f7b6714753c08a30ae21a471633e8567fee913875cae3ec204a87f91
+size 291876

requirements.txt

@@ -0,0 +1,6 @@
+pandas
+scikit-learn
+xgboost
+joblib
+shap
+matplotlib