app/gradio_simple.py

"""
HR Analytics - Simple Gradio App
Versi sederhana untuk quick deployment

Usage:
    python gradio_simple.py
"""

import gradio as gr
import pandas as pd
import pickle
import warnings
warnings.filterwarnings('ignore')


# Load model dan preprocessing objects
print("Loading model...")
with open('model/best_model_RF_SMOTETomek.pkl', 'rb') as f:
    model = pickle.load(f)
with open('model/scaler.pkl', 'rb') as f:
    scaler = pickle.load(f)
with open('model/label_encoders.pkl', 'rb') as f:
    encoders = pickle.load(f)
print("✓ Model loaded successfully!")


def predict_employee(tingkat_kepuasan, skor_evaluasi, jumlah_proyek,
                    jam_kerja_perbulan, lama_bekerja, kecelakaan_kerja,
                    promosi, divisi, gaji):
    """
    Predict resignation probability for an employee
    """
    
    # Create dataframe
    data = {
        'tingkat_kepuasan': [tingkat_kepuasan],
        'skor_evaluasi': [skor_evaluasi],
        'jumlah_proyek': [jumlah_proyek],
        'jam_kerja_perbulan': [jam_kerja_perbulan],
        'lama_bekerja': [lama_bekerja],
        'kecelakaan_kerja': [kecelakaan_kerja],
        'promosi': [promosi],
        'divisi': [divisi],
        'gaji': [gaji]
    }
    df = pd.DataFrame(data)
    
    # Encode categorical features
    for col in ['kecelakaan_kerja', 'promosi', 'divisi', 'gaji']:
        df[col] = encoders[col].transform(df[col])
    
    # Scale features
    X_scaled = scaler.transform(df)
    
    # Predict
    prediction = model.predict(X_scaled)[0]
    probability = model.predict_proba(X_scaled)[0]
    resign_prob = probability[1] * 100
    
    # Determine risk level
    if resign_prob < 30:
        risk_level = "🟢 LOW RISK"
        risk_color = "#2ecc71"
    elif resign_prob < 60:
        risk_level = "🟡 MEDIUM RISK"
        risk_color = "#f39c12"
    else:
        risk_level = "🔴 HIGH RISK"
        risk_color = "#e74c3c"
    
    # Result
    result = f"""
    ## Hasil Prediksi
    
    **Status:** {'AKAN RESIGN' if prediction == 1 else 'TIDAK AKAN RESIGN'}
    
    **Probabilitas Resign:** {resign_prob:.1f}%
    
    **Risk Level:** {risk_level}
    
    ---
    
    ### Informasi Karyawan:
    - Kepuasan: {tingkat_kepuasan:.2f}
    - Evaluasi: {skor_evaluasi:.2f}
    - Proyek: {jumlah_proyek}
    - Jam Kerja: {jam_kerja_perbulan} jam/bulan
    - Lama Kerja: {lama_bekerja} tahun
    - Divisi: {divisi}
    - Gaji: {gaji}
    """
    
    # Recommendations
    recs = ["### 💡 Rekomendasi:"]
    
    if resign_prob >= 60:
        recs.append("- ⚠️ URGENT: Schedule meeting segera")
        recs.append("- Review kompensasi dan benefit")
    
    if tingkat_kepuasan < 0.4:
        recs.append("- Tingkatkan kepuasan karyawan")
        recs.append("- Identifikasi sumber ketidakpuasan")
    
    if jam_kerja_perbulan > 250:
        recs.append("- Kurangi beban kerja")
        recs.append("- Improve work-life balance")
    
    if resign_prob < 30:
        recs.append("- ✅ Karyawan dalam kondisi baik")
        recs.append("- Maintain current engagement")
    
    return result, "\n".join(recs)


# Create Gradio interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    
    gr.Markdown("""
    # 🎯 HR Analytics - Prediksi Karyawan Resign
    
    Masukkan data karyawan untuk memprediksi kemungkinan resign
    """)
    
    with gr.Row():
        with gr.Column():
            gr.Markdown("### 📝 Input Data Karyawan")
            
            tingkat_kepuasan = gr.Slider(0, 1, value=0.5, step=0.01,
                                        label="Tingkat Kepuasan")
            skor_evaluasi = gr.Slider(0, 1, value=0.7, step=0.01,
                                     label="Skor Evaluasi")
            jumlah_proyek = gr.Slider(2, 7, value=3, step=1,
                                     label="Jumlah Proyek")
            jam_kerja_perbulan = gr.Slider(96, 310, value=200, step=1,
                                          label="Jam Kerja/Bulan")
            lama_bekerja = gr.Slider(2, 10, value=3, step=1,
                                    label="Lama Bekerja (tahun)")
            
            kecelakaan_kerja = gr.Radio(["tidak", "pernah"], value="tidak",
                                       label="Kecelakaan Kerja")
            promosi = gr.Radio(["tidak", "ya"], value="tidak",
                              label="Promosi (5 tahun terakhir)")
            
            divisi = gr.Dropdown(
                ["sales", "accounting", "hr", "technical", "support",
                 "management", "IT", "product_mng", "marketing", "RandD"],
                value="sales", label="Divisi"
            )
            
            gaji = gr.Radio(["low", "medium", "high"], value="medium",
                           label="Kategori Gaji")
            
            predict_btn = gr.Button("🔮 Prediksi", variant="primary", size="lg")
        
        with gr.Column():
            gr.Markdown("### 📊 Hasil Prediksi")
            output_result = gr.Markdown()
            output_recommendations = gr.Markdown()
    
    # Connect
    predict_btn.click(
        fn=predict_employee,
        inputs=[tingkat_kepuasan, skor_evaluasi, jumlah_proyek,
                jam_kerja_perbulan, lama_bekerja, kecelakaan_kerja,
                promosi, divisi, gaji],
        outputs=[output_result, output_recommendations]
    )
    
    gr.Markdown("""
    ---
    
    **Model:** Random Forest + SMOTE | **Akurasi:** 95%+
    """)


if __name__ == "__main__":
    demo.launch(share=True, debug=True)