****Here is a very simple code with PyCaret (execution output in attachment) :**** 

import pycaret 
from pycaret.regression import *

setup(data = train, target = 'SalePrice', preprocess=True, normalize=True, session_id=3141)

print(get_config('pipeline'))

best = compare_models(sort='RMSE', include = ['lightgbm'], n_select = 1)

predict_model(best);

#finalize_model refits a given model on the entire train dataset.
final_lightgbm = finalize_model(best)

pyCaret_predicted_test_y = predict_model(final_lightgbm, data=test)
pyCaret_predicted_test_y = pd.DataFrame(pyCaret_predicted_test_y['prediction_label'])

**Here is a code with sklearn which should be exactly equivalent and produce exactly the same result :**

from sklearn.model_selection import train_test_split, cross_validate
from sklearn.metrics import make_scorer, mean_squared_error, r2_score
from lightgbm import LGBMRegressor
import pandas as pd
import numpy as np
from scipy.stats import zscore

# Charger les données
train = pd.read_csv(r'C:\Perso\Data Science\house-prices-advanced-regression-techniques\train.csv',sep=',', encoding='latin-1')

# Prétraitement des données
object_feature_mask = train.dtypes == 'object'
object_cols = train.columns[object_feature_mask].tolist()

# Remplir les valeurs manquantes pour les variables catégorielles avec la valeur la plus fréquente
for col in object_cols:
   train[col] = train[col].astype('category')

# Conversion des types de données
int_cols = train.select_dtypes(include=['int']).columns.tolist()
for col in int_cols:
    train[col] = train[col].astype(np.int32)

numeric_cols = ['OverallQual', 'OverallCond', 'BsmtFullBath', 'BsmtHalfBath', 
                'FullBath', 'HalfBath', 'BedroomAbvGr', 'KitchenAbvGr', 
                'TotRmsAbvGrd', 'Fireplaces', 'GarageCars', 'MoSold']
for col in numeric_cols:
    train[col] = train[col].astype(np.int8)

float_cols = train.select_dtypes(include=['float']).columns.tolist()
for col in float_cols:
    train[col] = train[col].astype(np.float32)

# Séparation des features et de la target
X_train = train.drop(['SalePrice'], axis=1)
y_train = train['SalePrice']

numeric_columns = train.select_dtypes(include=['number']).columns.tolist()


# Normalisation des features numériques avec zscore
X_train[X_train.columns] = X_train.apply(lambda x: zscore(x) if x.name in numeric_columns else x)

# Division en ensemble d'entraînement et de test
X_train_train, X_test_train, y_train_train, y_test_train = train_test_split(X_train, y_train, train_size=0.7, random_state=3141)

# Initialisation du modèle
lgbm = LGBMRegressor()

# Création d'un scorer pour RMSE
rmse_scorer = make_scorer(mean_squared_error, squared=False)

# Évaluation du modèle avec cross_validate()
cv_results = cross_validate(
    lgbm, 
    X_train_train, 
    y_train_train, 
    scoring={'RMSE': rmse_scorer, 'R2': 'r2'}, 
    cv=10, 
    return_train_score=True,
    n_jobs=-1
)

# Affichage des résultats
print(f"RMSE TRAIN: {np.mean(cv_results['test_RMSE']):.4f}")
print(f"R2 TRAIN: {np.mean(cv_results['test_R2']):.4f}")




![Sklearn](https://github.com/user-attachments/assets/dc0bab30-c5bc-47bc-9aab-7557c6bbd46b)
![PyCaret](https://github.com/user-attachments/assets/c0632ac1-9c4a-496d-8cf4-c115e8913c89)