gridsearch_example for post

gridsearch_example.py

params_rf = {
    'clf__n_estimators':[
        100,
        300,
        500
    ],
    'clf__min_samples_leaf': [
        1, 
        2, 
        4
    ],
    'clf__min_samples_split': [
        2,
        5, 
        10
    ],
    'clf__max_features': [
        'auto', 
        'sqrt'
    ],
    'clf__bootstrap': [
        True, 
        False
    ],
    'clf__class_weight':[
        'balanced',
        'balanced_subsample',
        None,
        utils.compute_weights(y)
    ]
}

params_fit_rf = {
   
}

gs_rf = GridSearchCV(
    estimator=get_pipeline_model(RandomForestClassifier()),
    param_grid = params_rf,
    verbose=10, 
    scoring= {
        'accuracy': make_scorer(accuracy_score),
        'f1_score': make_scorer(f1_score, average='weighted'),
    }, 
    cv=StratifiedKFold(3), 
    return_train_score=True, 
    refit='accuracy', 
    n_jobs=-1
)

gs_rf = gs_rf.fit(X_train, y_train)

utils.py

def compute_weights(y):
    weights = compute_class_weight("balanced", np.unique(y), y)
    return { i:k for i,k in enumerate(weights)}