def run_logreg():
# CV
print()
recs, precs, accs = [], [], []
for i in range(len(cv_splits)):
print('CV Epoch : ' + str(i + 1))
cv_train, cv_test = train_test_split(cv_splits[i])
cv_train_X, cv_train_Y = get_X_Y(cv_train)
cv_test_X, cv_test_Y = get_X_Y(cv_test)
mlr_fit = mlr.fit(cv_train_X, cv_train_Y)
cv_pred = mlr.predict(cv_test_X)
# The Coefficients
print('Coefficients : \n', mlr.coef_)
# Recall Score
recall = r_s(cv_test_Y, cv_pred)
print('Recall Score : \n', recall)
# Precision Score
precision = p_s(cv_test_Y, cv_pred)
print('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(cv_test_Y, cv_pred)
print('Accuracy Score : \n', accuracy)
# Conusion Matix
print('Confusion Matrix : \n', c_m(cv_test_Y, cv_pred))
recs.append(recall)
precs.append(precision)
accs.append(accuracy)
print()
print('Average Recall Score : %f' % np.mean(recs))
print('Average Precision Score : %f' % np.mean(precs))
print('Average Accuracy Score : %f' % np.mean(accs))
print()
# Test
test_X, test_Y = get_X_Y(data_test)
test_pred = mlr.predict(test_X)
# The Coefficients
print('Test Coefficients : \n', mlr.coef_)
# Recall Score
recall = r_s(test_Y, test_pred)
print('Recall Score : \n', recall)
# Precision Score
precision = p_s(test_Y, test_pred)
print('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(test_Y, test_pred)
print('Accuracy Score : \n', accuracy)
# Conusion Matix
print('Confusion Matrix : \n', c_m(test_Y, test_pred))
print()
return None
def sgs():
# gaussian/rbf
print()
tg = time.time()
cs = [0.1, 0.5, 1.0, 2.0, 5.0]
sigmas = [0.1, 0.5, 1.0, 2.0, 4.0]
hyperparams = {'C': cs, 'gamma': sigmas}
rbf_svc = SVC(kernel='rbf', C=cs, gamma=sigmas, cache_size=4096)
rbf_svc_clf = GridSearchCV(rbf_svc, hyperparams, cv=5)
cv_train_X, cv_train_Y = get_X_Y(data_cv)
rbf_svc_fit = rbf_svc_clf.fit(cv_train_X, cv_train_Y)
rbf_svc_res = rbf_svc_clf.cv_results_
rbf_svc_params = rbf_svc_clf.best_params_
rbf_svc_score = rbf_svc_clf.best_score_
test_X, test_Y = get_X_Y(data_test)
test_pred = rbf_svc_clf.predict(test_X)
# The Coefficients
print('Test Estimator : \n', rbf_svc_clf.best_estimator_)
# Recall Score
recall = r_s(test_Y, test_pred)
print('Recall Score : \n', recall)
# Precision Score
precision = p_s(test_Y, test_pred)
print('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(test_Y, test_pred)
print('Accuracy Score : \n', accuracy)
# Conusion Matix
print('Confusion Matrix : \n', c_m(test_Y, test_pred))
tg = time.time() - tg
print('Time Secs : %f' % tg)
return None
def sls():
# linear
print()
tl = time.time()
cs = [0.1, 0.5, 1.0, 2.0, 5.0]
lin_svc = SVC(C=cs, kernel='linear', cache_size=4096)
hyperparams = {'C': cs}
lin_svc_clf = GridSearchCV(lin_svc, hyperparams, cv=5)
cv_train_X, cv_train_Y = get_X_Y(data_cv)
lin_svc_fit = lin_svc_clf.fit(cv_train_X, cv_train_Y)
lin_svc_res = lin_svc_clf.cv_results_
lin_svc_params = lin_svc_clf.best_params_
lin_svc_score = lin_svc_clf.best_score_
test_X, test_Y = get_X_Y(data_test)
test_pred = lin_svc_clf.predict(test_X)
# The Coefficients
print('Test Estimator : \n', lin_svc_clf.best_estimator_)
# Recall Score
recall = r_s(test_Y, test_pred)
print('Recall Score : \n', recall)
# Precision Score
precision = p_s(test_Y, test_pred)
print('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(test_Y, test_pred)
print('Accuracy Score : \n', accuracy)
# Conusion Matix
print('Confusion Matrix : \n', c_m(test_Y, test_pred))
tl = time.time() - tl
print('Time Secs : %f' % tl)
return None
def _train_all(names,
classifiers,
X,
y,
X_train,
X_test,
y_train,
y_test,
stats=True,
predict=""):
"""
Train each of the classifiers, and either output score or the predictions
"""
## ignore numpy warnings
from warnings import filterwarnings
filterwarnings('ignore')
## cycle around each classifier
classes = {1: "LIKELY", -1: "UNLIKELY"}
score = {1: 0, -1: 0}
trusts = {}
predictions = {}
for name, classifier in zip(names, classifiers):
## train each classifier
classifier.fit(X_train, y_train)
if stats == True:
_get_statistics(name, classifier, X, y, X_test, y_test)
if predict != "":
## Make prediction
prediction = classifier.predict(predict)[0]
## Increment counter for relevant score
score[prediction] += 1
predictions.update({name: prediction})
"""
reveal expected true negatives, false positives,
false negatives, true positives
"""
tn, fp, fn, tp = c_m(y_test, classifier.predict(X_test)).ravel()
## trust is the amount of time that the prediction was correct
trust_score = tp / (tp + fp) if prediction == 1 else tn / (tn + fn)
trust_score = round((trust_score * 100), 2)
trusts.update({name: trust_score})
if predict != "":
scores = pd.DataFrame({
'Recurrence': predictions,
'Confidence': trusts
})
pred_weight = scores.Recurrence * scores.Confidence
weights = pd.DataFrame({'Weights': pred_weight})
scores['Recurrence'] = scores['Recurrence'].apply(lambda x: classes[x])
print(scores)
classification = 1 if weights.Weights.mean() > 0 else -1
print(f"\nRecurrence judged {classes[classification]} at \
{round(abs(weights.Weights.mean()),2)} % confidence")
print(f"Poll of classifiers results:")
for index in score:
print(f"{classes[index]}: \t\t{score[index]}")
def kp():
# polynomial
print()
tp = time.time()
recs, precs, accs = [], [], []
alphas = [1.0]
degs = [2.0, 3.0] # M
hyperparams = {'alpha': alphas, 'degree': degs}
poly_krr = KernelRidge(kernel='poly',
alpha=alphas,
degree=degs,
gamma=1,
coef0=1)
poly_krr_clf = GridSearchCV(poly_krr, hyperparams, cv=5)
for batch in data_batches:
batch_train, batch_test = train_test_split(batch)
cv_train_X, cv_train_Y = get_X_Y(batch_train)
poly_krr_fit = poly_krr_clf.fit(cv_train_X, cv_train_Y)
poly_krr_res = poly_krr_clf.cv_results_
poly_krr_params = poly_krr_clf.best_params_
poly_krr_score = poly_krr_clf.best_score_
test_X, test_Y = get_X_Y(batch_test)
test_pred = poly_krr_clf.predict(test_X)
# The Coefficients
print('Test Estimator : \n', poly_krr_clf.best_estimator_)
# Recall Score
recall = r_s(test_Y, test_pred)
print('Recall Score : \n', recall)
# Precision Score
precision = p_s(test_Y, test_pred)
print('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(test_Y, test_pred)
print('Accuracy Score : \n', accuracy)
# Conusion Matix
print('Confusion Matrix : \n', c_m(test_Y, test_pred))
recs.append(recall)
precs.append(precision)
accs.append(accuracy)
print()
print('Average Test Recall Score : %f' % np.mean(recs))
print('Average Test Precision Score : %f' % np.mean(precs))
print('Average Test Accuracy Score : %f' % np.mean(accs))
tp = time.time() - tp
print('Time Secs : %f' % tp)
return None
def kg():
# gaussian/rbf
print()
tg = time.time()
recs, precs, accs = [], [], []
alphas = [1.0]
sigmas = [0.1, 0.5, 1.0, 2.0, 4.0]
hyperparams = {'alpha': alphas, 'gamma': sigmas}
rbf_krr = KernelRidge(kernel='rbf', alpha=alphas, gamma=sigmas)
rbf_krr_clf = GridSearchCV(rbf_krr, hyperparams, cv=5)
for batch in data_batches:
batch_train, batch_test = train_test_split(batch)
cv_train_X, cv_train_Y = get_X_Y(batch_train)
rbf_krr_fit = rbf_krr_clf.fit(cv_train_X, cv_train_Y)
rbf_krr_res = rbf_krr_clf.cv_results_
rbf_krr_params = rbf_krr_clf.best_params_
rbf_krr_score = rbf_krr_clf.best_score_
test_X, test_Y = get_X_Y(batch_test)
test_pred = rbf_krr_clf.predict(test_X)
# The Coefficients
print('Test Estimator : \n', rbf_krr_clf.best_estimator_)
# Recall Score
recall = r_s(test_Y, test_pred)
print('Recall Score : \n', recall)
# Precision Score
precision = p_s(test_Y, test_pred)
print('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(test_Y, test_pred)
print('Accuracy Score : \n', accuracy)
# Conusion Matix
print('Confusion Matrix : \n', c_m(test_Y, test_pred))
recs.append(recall)
precs.append(precision)
accs.append(accuracy)
print()
print('Average Test Recall Score : %f' % np.mean(recs))
print('Average Test Precision Score : %f' % np.mean(precs))
print('Average Test Accuracy Score : %f' % np.mean(accs))
tg = time.time() - tg
print('Time Secs : %f' % tg)
return None
def kl():
# linear
print()
tl = time.time()
recs, precs, accs = [], [], []
alphas = [1.0]
lin_krr = KernelRidge(alpha=alphas, kernel='linear')
hyperparams = {'alpha': alphas}
lin_krr_clf = GridSearchCV(lin_krr, hyperparams, cv=5)
for batch in data_batches:
batch_train, batch_test = train_test_split(batch)
cv_train_X, cv_train_Y = get_X_Y(batch_train)
lin_krr_fit = lin_krr_clf.fit(cv_train_X, cv_train_Y)
lin_krr_res = lin_krr_clf.cv_results_
lin_krr_params = lin_krr_clf.best_params_
lin_krr_score = lin_krr_clf.best_score_
test_X, test_Y = get_X_Y(batch_test)
test_pred = lin_krr_clf.predict(test_X)
# The Coefficients
print('Test Estimator : \n', lin_krr_clf.best_estimator_)
# Recall Score
recall = r_s(test_Y, test_pred)
print('Recall Score : \n', recall)
# Precision Score
precision = p_s(test_Y, test_pred)
print('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(test_Y, test_pred)
print('Accuracy Score : \n', accuracy)
# Conusion Matix
print('Confusion Matrix : \n', c_m(test_Y, test_pred))
recs.append(recall)
precs.append(precision)
accs.append(accuracy)
print()
print('Average Test Recall Score : %f' % np.mean(recs))
print('Average Test Precision Score : %f' % np.mean(precs))
print('Average Test Accuracy Score : %f' % np.mean(accs))
tl = time.time() - tl
print('Time Secs : %f' % tl)
return None
def sps():
# polynomial
print()
tp = time.time()
cs = [0.1, 0.5, 1.0, 2.0, 5.0]
degs = [2.0, 3.0] # M
hyperparams = {'C': cs, 'degree': degs}
poly_svc = SVC(kernel='poly',
C=cs,
degree=degs,
gamma=1,
coef0=1,
cache_size=4096)
poly_svc_clf = GridSearchCV(poly_svc, hyperparams, cv=5)
cv_train_X, cv_train_Y = get_X_Y(data_cv)
poly_svc_fit = poly_svc_clf.fit(cv_train_X, cv_train_Y)
poly_svc_res = poly_svc_clf.cv_results_
poly_svc_params = poly_svc_clf.best_params_
poly_svc_score = poly_svc_clf.best_score_
test_X, test_Y = get_X_Y(data_test)
test_pred = poly_svc_clf.predict(test_X)
# The Coefficients
print('Test Estimator : \n', poly_svc_clf.best_estimator_)
# Recall Score
recall = r_s(test_Y, test_pred)
print('Recall Score : \n', recall)
# Precision Score
precision = p_s(test_Y, test_pred)
print('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(test_Y, test_pred)
print('Accuracy Score : \n', accuracy)
# Conusion Matix
print('Confusion Matrix : \n', c_m(test_Y, test_pred))
tp = time.time() - tp
print('Time Secs : %f' % tp)
return None
print ('Time Taken To Test : %f Secs.' % tl)
# Round Test Predictions to Avoid Multiclass Continuous Targets Error
test_pred = np.round(test_pred)
# The Best Estimator
print ('Test Estimator : \n', lin_krr_clf.best_estimator_)
# Recall Score
recall = r_s(test_Y, test_pred, average='micro')
print ('Recall Score : \n', recall)
# Precision Score
precision = p_s(test_Y, test_pred, average='micro')
print ('Precision Score : \n', precision)
# Accuracy Score
accuracy = a_s(test_Y, test_pred)
print ('Accuracy Score : \n', accuracy)
# Confusion Matix
print('Confusion Matrix : \n', c_m(test_Y, test_pred))
print ()
# Polynomial Kernel Ridge Regression
print ()
# Functions and Parameters
alphas = [1.0]
degs = [2.0, 3.0, 4.0] # M
hyperparams = {'alpha' : alphas, 'degree' : degs}
poly_krr = KernelRidge(kernel='poly', alpha=alphas, degree=degs, gamma=1, coef0=1)
# Polynomial KRR Initializer
poly_krr_clf = GridSearchCV(poly_krr, hyperparams, cv=5)
# Train
tp = time.time()
poly_krr_fit = poly_krr_clf.fit(cv_train_X, cv_train_Y)