def dataset3_svc_sgd(dataset):
x_original = dataset.loc[:, dataset.columns != 'labels']
labels = dataset['labels']
fselect = FeatureSelection(dataset, x_original, labels)
model_lsvc = LinearSVC(C=0.1, penalty="l1", dual=False)
#model= linearsvs
X_fit_model, X_transf_model,column_selected,feature_importances,feature_importances_DF,dataset_features= \
fselect.select_from_model_feature_elimination(model=model_lsvc)
ml = MachineLearning(X_transf_model, labels, classes=['pos', 'neg'])
filename = 'crmapp/ml_models/dataset3_all_'
print('best model sgd')
best_sgd_model = ml.train_best_model('sgd')
print('score_test_set_sgd')
print(ml.score_testset(best_sgd_model))
print('roc curve')
ml.plot_roc_curve(best_sgd_model)
pickle.dump(best_sgd_model, open(filename + 'sgd_model_15092020.sav',
'wb'))
def dataset3_univariate_svm(dataset):
x_original = dataset.loc[:, dataset.columns != 'labels']
labels = dataset['labels']
fselect = FeatureSelection(dataset, x_original, labels)
model_lsvc = LinearSVC(C=0.1, penalty="l1", dual=False)
# #**Select KBest**
# #KBest com *mutual info classif*
X_fit_univariate, X_transf_univariate,column_selected,scores,dataset_features = \
fselect.univariate(score_func=mutual_info_classif, mode='k_best', param=500)
ml = MachineLearning(X_transf_univariate, labels, classes=['pos', 'neg'])
filename = 'crmapp/ml_models/dataset3_all_'
#tests models
print('best model svm')
best_svm_model = ml.train_best_model('svm')
print('score_test_set_svm')
print(ml.score_testset(best_svm_model))
print('roc curve')
ml.plot_roc_curve(best_svm_model)
def dataset1_tree_svm(dataset):
x_original = dataset.loc[:, dataset.columns != 'labels']
labels = dataset['labels']
fselect = FeatureSelection(dataset, x_original, labels)
model_tree = ExtraTreesClassifier(n_estimators=50)
# Select from model
#model= Tree classifier. 50 estiamtors
X_fit_model, X_transf_model,column_selected,feature_importances,feature_importances_DF,dataset_features= \
fselect.select_from_model_feature_elimination(model=model_tree)
ml = MachineLearning(X_transf_model, labels, classes=['pos', 'neg'])
filename = 'crmapp/ml_models/dataset1_all_'
#tests models
print('best model svm')
best_svm_model = ml.train_best_model('svm')
print('score_test_set_svm')
print(ml.score_testset(best_svm_model))
print('roc curve')
ml.plot_roc_curve(best_svm_model)
pickle.dump(best_svm_model, open(filename + 'svm_model_15092020.sav',
'wb'))
def predictions(model,
dataset_file='dataset2_usar.csv',
dataset='descriptors.sav',
gap=1):
dataset_prev = read_dataset(dataset_file)
dataset_prev_X = dataset_prev.loc[:, 'seq']
dataset_prev_y = dataset_prev.loc[:, 'labels']
model = pickle.load(open(model, 'rb'))
# dataset = test_preprocess(get_dataset(dataset))
dataset = get_dataset(dataset)
x_original = dataset.loc[:, dataset.columns != 'labels']
labels = dataset.loc[:, 'labels']
ml = MachineLearning(x_original, labels, classes=['pos', 'neg'])
"""
results = {'total':len(dataset_prev_X),
'true_pos':0,
'true_neg':0,
'false_pos':0,
'false_neg':0}
"""
prevs = {}
for i in range(len(dataset_prev_X)):
seq = dataset_prev_X[i]
print(i, seq)
current_label = dataset_prev_y[i]
result = ml.predict_window(model,
seq=seq,
x=None,
window_size=len(seq),
gap=gap,
features=[],
names=None,
y=None,
filename=None)
#df = ml.predict(model, x=X_test, seqs=y_test)
prevs[i] = (result, current_label)
"""
if result['probability'][0] > 0.5 and current_label == 'pos':
results['true_pos'] += 1
elif result['probability'][0] <= 0.5 and current_label == 'neg':
results['true_neg'] += 1
elif result['probability'][0] <= 0.5 and current_label == 'pos':
results['false_neg'] += 1
elif result['probability'][0] > 0.5 and current_label == 'neg':
results['false_pos'] += 1
"""
return prevs
def test_machine_learning(dataset):
#split dataset
# dataset = pd.read_csv(r'datasets/dataset1_test_clean_fselection.csv', delimiter=',')
print(dataset.shape)
x_original = dataset.loc[:, dataset.columns != 'labels']
labels = dataset['labels']
#create Machine learning object
ml = MachineLearning(x_original, labels, classes=['pos', 'neg'])
filename = 'crmapp/ml_models/dataset3_all_'
#tests models
print('best model svm')
best_svm_model = ml.train_best_model('svm')
pickle.dump(best_svm_model, open(filename + 'svm_model_24082020.sav',
'wb'))
print('best model rf')
best_rf_model = ml.train_best_model('rf')
pickle.dump(best_rf_model, open(filename + 'rf_model_24082020.sav', 'wb'))
print('best model sgd')
best_sgd_model = ml.train_best_model('sgd')
pickle.dump(best_sgd_model, open(filename + 'sgd_model_24082020.sav',
'wb'))
print('best model gradient boosting')
best_gboosting_model = ml.train_best_model('gboosting')
pickle.dump(best_gboosting_model,
open(filename + 'gboosting_model_24082020.sav', 'wb'))
print('best model lr')
best_lr_model = ml.train_best_model('lr')
pickle.dump(best_lr_model, open(filename + 'lr_model_24082020.sav', 'wb'))
# feature importance of models
ml.features_importances(best_svm_model, 'svm')
ml.features_importances(best_rf_model, 'rf')
# ml.features_importances(best_sgd_model,'sgd')
# ml.features_importances(best_gboosting_model,'gboosting')
# ml.features_importances(best_lr_model,'lr')
print('best model nn')
best_nn_model = ml.train_best_model('nn')
pickle.dump(best_nn_model, open(filename + 'nn_model_24082020.sav', 'wb'))
print('best model gnb')
best_gnb_model = ml.train_best_model('gnb')
pickle.dump(best_gnb_model, open(filename + 'gnb_model_24082020.sav',
'wb'))
print('best model knn')
best_knn_model = ml.train_best_model('knn')
pickle.dump(best_knn_model, open(filename + 'knn_model_24082020.sav',
'wb'))
#plot validation curve
print('plot validation_svm')
ml.plot_validation_curve(
best_svm_model,
param_name='clf__C',
param_range=[0.00001, 0.0001, 0.001, 0.01, 0.1, 1, 10, 100])
# print('plot validation_gboosting')
# ml.plot_validation_curve(best_gboosting_model, param_name='clf__n_estimators',
# param_range=[ 1, 10,100,500])
print('score_test_set_svm')
print(ml.score_testset(best_svm_model))
print('score_test_set_rf')
print(ml.score_testset(best_rf_model))
print('score_test_set_rf')
print(ml.score_testset(best_rf_model))
print('score_test_set_gboosting')
print(ml.score_testset(best_gboosting_model))
print('score_test_set_lr')
print(ml.score_testset(best_lr_model))