ML_AI_training/UQ_RF.py

140 lines
4.6 KiB
Python
Executable file

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Wed May 18 06:03:24 2022
@author: tanu
"""
#%% RandomForest + hyperparam: BaseEstimator: ClfSwitcher()
class ClfSwitcher(BaseEstimator):
def __init__(
self,
estimator = SGDClassifier(),
):
"""
A Custom BaseEstimator that can switch between classifiers.
:param estimator: sklearn object - The classifier
"""
self.estimator = estimator
def fit(self, X, y=None, **kwargs):
self.estimator.fit(X, y)
return self
def predict(self, X, y=None):
return self.estimator.predict(X)
def predict_proba(self, X):
return self.estimator.predict_proba(X)
def score(self, X, y):
return self.estimator.score(X, y)
parameters = [
{
'clf__estimator': [RandomForestClassifier(**rs
, **njobs
, bootstrap = True
, oob_score = True)],
'clf__estimator__max_depth': [4, 6, 8, 10, 12, 16, 20, None]
, 'clf__estimator__class_weight':['balanced','balanced_subsample']
, 'clf__estimator__n_estimators': [10, 25, 50, 100]
, 'clf__estimator__criterion': ['gini', 'entropy', 'log_loss']
, 'clf__estimator__max_features': ['sqrt', 'log2', None] #deafult is sqrt
, 'clf__estimator__min_samples_leaf': [1, 2, 3, 4, 5, 10]
, 'clf__estimator__min_samples_split': [2, 5, 15, 20]
}
]
# Create pipeline
pipeline = Pipeline([
('pre', MinMaxScaler()),
('clf', ClfSwitcher()),
])
# Grid search i.e hyperparameter tuning and refitting on mcc
gscv_rf = GridSearchCV(pipeline
, parameters
#, scoring = 'f1', refit = 'f1'
, scoring = mcc_score_fn, refit = 'mcc'
, cv = skf_cv
, **njobs
, return_train_score = False
, verbose = 3)
# Fit
gscv_rf_fit = gscv_rf.fit(X, y)
gscv_rf_fit_be_mod = gscv_rf_fit.best_params_
gscv_rf_fit_be_res = gscv_rf_fit.cv_results_
print('Best model:\n', gscv_rf_fit_be_mod)
print('Best models score:\n', gscv_rf_fit.best_score_, ':' , round(gscv_rf_fit.best_score_, 2))
print('\nMean test score from fit results:', round(mean(gscv_rf_fit_be_re['mean_test_mcc']),2))
print('\nMean test score from fit results:', round(np.nanmean(gscv_rf_fit_be_res['mean_test_mcc']),2))
######################################
# Blind test
######################################
# See how it does on the BLIND test
#print('\nBlind test score, mcc:', )
test_predict = gscv_rf_fit.predict(X_bts)
print(test_predict)
print(np.array(y_bts))
y_btsf = np.array(y_bts)
print(accuracy_score(y_btsf, test_predict))
print(matthews_corrcoef(y_btsf, test_predict))
# create a dict with all scores
rf_bts_dict = {#'best_model': list(gscv_rf_fit_be_mod.items())
'bts_fscore' : None
, 'bts_mcc' : None
, 'bts_precision': None
, 'bts_recall' : None
, 'bts_accuracy' : None
, 'bts_roc_auc' : None
, 'bts_jaccard' : None }
rf_bts_dict
rf_bts_dict['bts_fscore'] = round(f1_score(y_bts, test_predict),2)
rf_bts_dict['bts_mcc'] = round(matthews_corrcoef(y_bts, test_predict),2)
rf_bts_dict['bts_precision'] = round(precision_score(y_bts, test_predict),2)
rf_bts_dict['bts_recall'] = round(recall_score(y_bts, test_predict),2)
rf_bts_dict['bts_accuracy'] = round(accuracy_score(y_bts, test_predict),2)
rf_bts_dict['bts_roc_auc'] = round(roc_auc_score(y_bts, test_predict),2)
rf_bts_dict['bts_jaccard'] = round(jaccard_score(y_bts, test_predict),2)
rf_bts_dict
# Create a df from dict with all scores
pd.DataFrame.from_dict(rf_bts_dict, orient = 'index', columns = 'best_model')
rf_bts_df = pd.DataFrame.from_dict(rf_bts_dict,orient = 'index')
rf_bts_df.columns = ['Logistic_Regression']
print(rf_bts_df)
# Create df with best model params
model_params = pd.Series(['best_model_params', list(gscv_rf_fit_be_mod.items() )])
model_params_df = model_params.to_frame()
model_params_df
model_params_df.columns = ['Logistic_Regression']
model_params_df.columns
# Combine the df of scores and the best model params
rf_bts_df.columns
rf_output = pd.concat([model_params_df, rf_bts_df], axis = 0)
rf_output
# Format the combined df
# Drop the best_model_params row from rf_output
rf_df = rf_output.drop([0], axis = 0)
rf_df
#FIXME: tidy the index of the formatted df
###############################################################################