prediksi_model_inverse.py

import os
import streamlit as st
import pandas as pd
import numpy as np
import joblib
import os
from datetime import datetime
from sklearn.preprocessing import MinMaxScaler
import openpyxl
from scipy.optimize import differential_evolution

def predict_forward_from_params(product_name: str, params_dict: dict, model_folder: str) -> float:
    """
    Jalankan forward modelling XGBoost many-to-one untuk 1 baris input parameter.
    Dipakai baik untuk Validasi maupun Simulasi agar konsisten.
    """
    model_filename = f"model_checkpoint_xgb_{product_name}.joblib"
    model_path = os.path.join(model_folder, model_filename)

    if not os.path.exists(model_path):
        raise FileNotFoundError(f"File model untuk produk {product_name} tidak ditemukan: {model_path}")

    artifacts = joblib.load(model_path)

    # Struktur bundle: dict dengan kunci 'model', 'poly_transformer', 'input_features', 'poly_feature_names'
    if isinstance(artifacts, dict):
        fwd_model = artifacts.get("model", artifacts)
        poly_transformer = artifacts.get("poly_transformer", None)
        input_features = artifacts.get("input_features", list(params_dict.keys()))
        poly_feature_names = artifacts.get("poly_feature_names", None)
    else:
        # Fallback: kalau bukan dict, anggap langsung model
        fwd_model = artifacts
        poly_transformer = None
        input_features = list(params_dict.keys())
        poly_feature_names = None

    # Susun DataFrame satu baris
    X_base = pd.DataFrame([params_dict])

    # Pastikan semua fitur ada
    missing = [f for f in input_features if f not in X_base.columns]
    if missing:
        raise ValueError(
            "Fitur berikut dibutuhkan oleh model namun tidak ada di input: "
            + ", ".join(missing)
        )

    # Urutkan kolom sesuai urutan training
    X_base = X_base[input_features]

    # Polynomial features jika ada
    if poly_transformer is not None:
        X_poly = poly_transformer.transform(X_base)
        if (poly_feature_names is not None) and (len(poly_feature_names) == X_poly.shape[1]):
            X_final = pd.DataFrame(X_poly, columns=poly_feature_names)
        else:
            X_final = pd.DataFrame(X_poly)
    else:
        X_final = X_base

    # Prediksi
    y_pred = fwd_model.predict(X_final)[0]
    return float(y_pred)