Notes

from sklearn.model_selection import train_test_split, GridSearchCV
from sklearn.linear_model import SGDClassifier
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier


# Define the hyper-parameter grids
param_grid_sgd = {
'classifier__loss': ['hinge', 'log_loss', 'modified_huber'],
'classifier__alpha': [0.0001, 0.001, 0.01],
'classifier__penalty': ['l2', 'l1', 'elasticnet']
}

param_grid_svm = {
'classifier__C': [0.1, 1, 10, 100],
'classifier__gamma': [1, 0.1, 0.01, 0.001],
'classifier__kernel': ['rbf', 'linear']
}

param_grid_rf = {
'classifier__n_estimators': [10, 50, 100],
'classifier__max_depth': [None, 10, 20, 30],
'classifier__min_samples_split': [2, 5, 10],
'classifier__min_samples_leaf': [1, 2, 4]
}

# Create pipelines for each model

pipeline_svm = Pipeline(steps=[('preprocessor', preprocessor),
('classifier', SVC(random_state=92))])

pipeline_rf = Pipeline(steps=[('preprocessor', preprocessor),
('classifier',
RandomForestClassifier(random_state=92))])

# Use GridSearchCV for hyper-parameter optimization
grid_sgd = GridSearchCV(pipeline_sgd, param_grid_sgd, cv=cv,
scoring='f1',refit=True,n_jobs=-1)
grid_svm = GridSearchCV(pipeline_svm, param_grid_svm, cv=cv,
scoring='f1',refit=True,n_jobs=-1)
grid_rf = GridSearchCV(pipeline_rf, param_grid_rf, cv=cv,
scoring='f1',refit=True,n_jobs=-1)

# Fit the models
grid_sgd.fit(X_train, y_train)
grid_svm.fit(X_train, y_train)
grid_rf.fit(X_train, y_train)

# Get the best models
best_sgd = grid_sgd.best_estimator_
best_svm = grid_svm.best_estimator_
best_rf = grid_rf.best_estimator_

# Evaluate each model on the test set
models = {
'SGD Classifier': best_sgd,
'SVM': best_svm,
'Random Forest': best_rf
}