def run_random_forest(X_train, y_train, X_test, y_test):
print("Fitting Random Forest")
X_train = dp.factorize_variables(X_train)
X_test = dp.factorize_variables(X_test)
max_features = [None, 'auto', 'log2']
params = {'criterion': ['gini'],
'random_state': [1234],
'n_estimators': [100, 200],
'max_features': max_features,
'oob_score': [False],
# EOSL -- just fit max depth trees. not going to overfit
'max_depth': [None, 10],
'n_jobs': [1],
}
cv_func, y = dp.get_kfold_obj(y_train, k = 3)
grid = GridSearchCV(RandomForestClassifier(), params, cv=cv_func,
verbose=2)
grid.fit(X_train, y_train.values)
print(grid.best_score_)
print(grid.best_estimator_)
print("Training set score {}".format(grid.score(X_train,y_train)))
print("Test set score {}".format(grid.score(X_test,y_test)))
return grid
y_train = pd.Series(psf.adjust_y(y_train))
y_test= pd.Series(psf.adjust_y(y_test))
X_ID = train.pop('Id')
Y_ID = test.pop('Id')
# are there any missing values?
dp.print_columns_with_missing(X_train)
dp.print_columns_with_missing(X_test)
# Run the model
rf_model = rrf.run_random_forest(X_train, y_train, X_test, y_test)
print(rf_model.get_params())
rf_final_model = rf_model.fit(X_data_full, y_data_full.values)
# Score the actual test set
test = dp.read_data('{}data/test.csv'.format(file_path))
test = test.set_index('Id')
test = dp.factorize_variables(test)
test_predictions = rf_final_model.predict(test)
test = pd.DataFrame(np.transpose([test.index, test_predictions]))
test.columns = ["Id", "Hazard"]
# Store results and pickle model
test.to_csv('{}Output/{}.csv'.format(file_path, model_name), drop = True)
#with open('/Users/Adrianna/Desktop/Kaggle/Liberty/Output/rf.pkl', 'wb') as f:
# cPickle.dump(rf_model, f)
