import streamlit as st
import pandas as pd
import numpy as np
from catboost import CatBoostClassifier
import joblib
import shap
import matplotlib.pyplot as plt
import os

# --- KONFIGURASI HALAMAN ---
st.set_page_config(
    page_title="EWS Prediksi Dropout",
    page_icon="🎓",
    layout="wide"
)

# --- FUNGSI LOAD MODEL ---
@st.cache_resource
def load_resources():
    current_dir = os.path.dirname(os.path.abspath(__file__))
    model_path = os.path.join(current_dir, "catboost_dropout_model.cbm")
    scaler_path = os.path.join(current_dir, "scaler.pkl")

    if not os.path.exists(model_path):
        st.error(f"❌ File model tidak ditemukan di: {model_path}")
        st.stop()

    model = CatBoostClassifier()
    model.load_model(model_path)
    scaler = joblib.load(scaler_path)
    return model, scaler

try:
    model, scaler = load_resources()
except Exception as e:
    st.error(f"Error loading resources: {e}")
    st.stop()

# --- JUDUL ---
st.title("🎓 Early Warning System: Student Dropout Prediction")
st.markdown("Aplikasi ini menggunakan **Machine Learning (CatBoost)** untuk mendeteksi dini mahasiswa yang berisiko *dropout*.")

# --- SIDEBAR INPUT ---
st.sidebar.header("📝 Input Data Mahasiswa")

def user_input_features():
    st.sidebar.subheader("1. Data Administratif")
    jalur_masuk = st.sidebar.selectbox("Jalur Masuk (Kode)", options=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14]) 
    provinsi = st.sidebar.number_input("Kode Provinsi", min_value=0, max_value=50, value=35)
    kurikulum = st.sidebar.selectbox("Kurikulum", options=[0, 1, 2, 3])
    
    st.sidebar.subheader("2. Riwayat Akademik")
    ips1 = st.sidebar.slider("IPS Semester 1", 0.0, 4.0, 3.5)
    ips2 = st.sidebar.slider("IPS Semester 2", 0.0, 4.0, 3.4)
    ips3 = st.sidebar.slider("IPS Semester 3", 0.0, 4.0, 3.2)
    ips4 = st.sidebar.slider("IPS Semester 4", 0.0, 4.0, 3.0)
    
    nilai1 = st.sidebar.number_input("Nilai Angka Sem 1", 0.0, 100.0, 80.0)
    nilai2 = st.sidebar.number_input("Nilai Angka Sem 2", 0.0, 100.0, 78.0)
    nilai3 = st.sidebar.number_input("Nilai Angka Sem 3", 0.0, 100.0, 75.0)
    nilai4 = st.sidebar.number_input("Nilai Angka Sem 4", 0.0, 100.0, 70.0)
    
    status1 = st.sidebar.selectbox("Status Sem 1 (1=Aktif)", [0, 1], index=1)
    status2 = st.sidebar.selectbox("Status Sem 2 (1=Aktif)", [0, 1], index=1)
    status3 = st.sidebar.selectbox("Status Sem 3 (1=Aktif)", [0, 1], index=1)
    status4 = st.sidebar.selectbox("Status Sem 4 (1=Aktif)", [0, 1], index=1)
    
    st.sidebar.subheader("3. Beban Studi")
    jumlah_mk = st.sidebar.number_input("Total MK Diambil", min_value=10, max_value=150, value=40)
    banyak_mk_ulang = st.sidebar.number_input("Total MK Mengulang", min_value=0, max_value=50, value=0)

    # Feature Engineering
    delta_ips_1_4 = ips4 - ips1
    delta_ips_3_4 = ips4 - ips3
    rata_rata_total = (nilai1 + nilai2 + nilai3 + nilai4) / 4
    rasio_mengulang = banyak_mk_ulang / (jumlah_mk + 1e-5)

    data = {
        'provinsi': provinsi,
        'jalur masuk': jalur_masuk,
        'kurikulum': kurikulum,
        'jumlah mk diambil': jumlah_mk,
        'banyak mk mengulang': banyak_mk_ulang,
        'rata rata nilai semester 1': nilai1,
        'rata rata nilai semester 2': nilai2,
        'rata rata nilai semester 3': nilai3,
        'rata rata nilai semester 4': nilai4,
        'ips semester 1': ips1,
        'ips semester 2': ips2,
        'ips semester 3': ips3,
        'ips semester 4': ips4,
        'status semester 1': status1,
        'status semester 2': status2,
        'status semester 3': status3,
        'status semester 4': status4,
        'delta_ips_1_4': delta_ips_1_4,
        'delta_ips_3_4': delta_ips_3_4,
        'rata_rata_total': rata_rata_total,
        'rasio_mengulang': rasio_mengulang
    }
    return pd.DataFrame(data, index=[0])

input_df = user_input_features()

# --- MAIN LAYOUT ---
col1, col2 = st.columns([1, 2])

with col1:
    st.info("Pastikan data di sidebar sudah benar sebelum menekan tombol analisis.")
    st.write("**Preview Data Input:**")
    st.dataframe(input_df.T, height=400)

with col2:
    if st.button("🚀 Analisis Risiko Dropout", type="primary"):
        with st.spinner('Sedang menganalisis...'):
            try:
                # 1. Transform Data
                input_scaled = scaler.transform(input_df)
                
                # 2. Prediksi
                prediction = model.predict(input_scaled)
                proba = model.predict_proba(input_scaled)[0][1]
                
                # 3. Tampilkan Hasil
                st.subheader("Hasil Analisis")
                
                if prediction[0] == 1:
                    st.error(f"⚠️ **STATUS: BERISIKO DROPOUT**")
                    st.metric("Probabilitas Dropout", f"{proba*100:.1f}%")
                    st.warning("Mahasiswa ini memerlukan perhatian akademik khusus.")
                else:
                    st.success(f"✅ **STATUS: AMAN (Pass)**")
                    st.metric("Probabilitas Dropout", f"{proba*100:.1f}%")
                    st.info("Performa akademik mahasiswa terpantau stabil.")

                # 4. Visualisasi SHAP (Waterfall Plot)
                st.markdown("---")
                st.subheader("📊 Faktor Penentu Keputusan")
                st.caption("Grafik ini menunjukkan fitur apa yang paling mendorong (+) atau mengurangi (-) risiko dropout.")
                
                explainer = shap.TreeExplainer(model)
                shap_values = explainer.shap_values(input_scaled)
                
                # Menggunakan Waterfall Plot (Lebih stabil & informatif)
                fig, ax = plt.subplots(figsize=(8, 6))
                shap.waterfall_plot(
                    shap.Explanation(
                        values=shap_values[0], 
                        base_values=explainer.expected_value, 
                        data=input_df.iloc[0],
                        feature_names=input_df.columns
                    ),
                    max_display=10,
                    show=False
                )
                st.pyplot(fig) # Menggambar figure waterfall
                
            except Exception as e:
                st.error(f"Terjadi kesalahan: {e}")

# --- FOOTER ---
st.markdown("---")
st.markdown("© 2025 EWS System | Institut Teknologi Sepuluh Nopember")