XGBoost Pipeline with ColumnTransformer: Handling categorical + numerical data and improving accuracy

from sklearn.modelselection import GridSearchCV
Define hyperparameter grid
paramgrid = {
    'classifiernestimators': [100, 200],
    'classifiermaxdepth': [3, 5, 7],
    'classifierlearningrate': [0.01, 0.1],
    'classifiersubsample': [0.8, 1.0],
    'classifiercolsamplebytree': [0.8, 1.0]
}
Initialize GridSearchCV
gridsearch = GridSearchCV(
    estimator=model,          # your existing pipeline
    paramgrid=paramgrid,
    cv=5,
    scoring='accuracy',
    njobs=-1,
    verbose=1
)
Fit grid search on training data
print("Running Grid Search...")
gridsearch.fit(Xtrain, ytrain)
Get best model
bestmodel = gridsearch.bestestimator
print("Best Parameters:", gridsearch.bestparams)