###############################################################################
#                     7. Tuning a classifier to use with RFECV                #
###############################################################################
# Define classifier to use as the base of the recursive feature elimination algorithm
selected_classifier = "Random Forest"
classifier = classifiers[selected_classifier]

# Tune classifier (Took = 4.8 minutes)

# Scale features via Z-score normalization
scaler = StandardScaler()

# Define steps in pipeline
steps = [("scaler", scaler), ("classifier", classifier)]

# Initialize Pipeline object
pipeline = Pipeline(steps = steps)

# Define parameter grid
param_grid = parameters[selected_classifier]

# Initialize GridSearch object
gscv = GridSearchCV(pipeline, param_grid, cv = 5,  n_jobs= -1, verbose = 1, scoring = "roc_auc")

# Fit gscv
print(f"Now tuning {selected_classifier}. Go grab a beer or something.")
gscv.fit(X_train, np.ravel(y_train))

# Get best parameters and score
best_params = gscv.best_params_
best_score = gscv.best_score_

# Update classifier parameters
tuned_params = {item[12:]: best_params[item] for item in best_params}
classifier.set_params(**tuned_params)