# %% [1] IMPORT LIBRARIES & LOAD DATA
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report, accuracy_score
import joblib

print("--- [STEP 3] Pelatihan Model Random Forest 30 Jurusan ---")
print("Memuat dataset yang sudah diproses...")
df = pd.read_csv('dataset_siswa_siap_model.csv')

# %% [2] MEMISAHKAN FITUR DAN TARGET (DATA SPLIT)
print("Membagi fitur dan target...")
X = df.drop('rekomendasi_jurusan', axis=1)
y = df['rekomendasi_jurusan']

print("Membagi data menjadi data latih (80%) dan data uji (20%)...")
# Karena ada 30 kelas, stratify sangat wajib agar distribusi per kelas merata
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y)
print(f"Jumlah Data Latih: {X_train.shape[0]} | Jumlah Data Uji: {X_test.shape[0]}")

# %% [3] TRAINING MODEL RANDOM FOREST (HYPERPARAMETER ANTI-OVERFIT)
print("\nMelatih model Random Forest...")
# Dengan data dimensi RIASEC, kita bisa membebaskan pohon sedikit lebih dalam
rf_model = RandomForestClassifier(
    n_estimators=200,       # 200 pohon sudah sangat cukup untuk 30 jurusan
    max_depth=12,           # Cegah menghafal terlalu dalam (Anti Overfitting)
    min_samples_split=4,    # Split akan terjadi jika minimal ada 4 sampel
    min_samples_leaf=2,     # Daun terakhir minimal berisi 2 sampel
    max_features='sqrt',
    class_weight='balanced',
    random_state=42,
    n_jobs=-1
)

rf_model.fit(X_train, y_train)
print("Proses pelatihan selesai!")

# %% [4] EVALUASI MODEL
print("\nMengevaluasi performa model pada data uji...")
y_pred = rf_model.predict(X_test)

akurasi = accuracy_score(y_test, y_pred)
print(f"\n======================================")
print(f"AKURASI MODEL 30 JURUSAN: {akurasi * 100:.2f}%")
print(f"======================================")

print("\nLaporan Detail Prediksi Keseluruhan:")
print(classification_report(y_test, y_pred))

# Cek Feature Importances (Membuktikan bahwa RIASEC akan memiliki bobot besar)
importances = rf_model.feature_importances_
feature_names = X.columns
feature_imp_df = pd.DataFrame({'Fitur': feature_names, 'Bobot': importances}).sort_values('Bobot', ascending=False)

plt.figure(figsize=(10, 8))
sns.barplot(x='Bobot', y='Fitur', data=feature_imp_df, palette='magma')
plt.title('Tingkat Kepentingan Fitur (Feature Importances)', fontsize=14)
plt.tight_layout()
plt.show()

# %% [5] SIMPAN MODEL
print("\nMenyimpan struktur model yang sudah berhasil dilatih...")
joblib.dump(rf_model, 'model_rekomendasi_15mapel.pkl')
print("Model berhasil diamankan ke dalam format 'model_rekomendasi_15mapel.pkl'!")
# %%