def run_model():
# Load data
X_train_scaled, y_train, X_test_scaled, y_test = clean_data(scaled=1)
# Build the naive bayes models
gaussianNB = GaussianNB()
bernoulliNB = BernoulliNB()
multinomialNB = MultinomialNB()
for nb in [gaussianNB, bernoulliNB, multinomialNB]:
print(f'Working {str(nb)}\n')
nb.fit(X_train_scaled, y_train)
y_pred = nb.predict(X_test_scaled)
report = classification_report(y_true=y_test, y_pred=y_pred)
print(report)
print(confusion_matrix(y_true=y_test, y_pred=y_pred))
# Confusion matrix
CM = confusion_matrix(y_true=y_test, y_pred=y_pred, labels=[0, 1])
cm(cm=CM,
target_names=['Burpee no jump', 'Burpee'],
title=str(nb) + ' CM')
# ROC
ROC(Model=nb, Y_test=y_test, X_test=X_test_scaled)
def run_model():
# Load data
X_train_scaled, y_train, X_test_scaled, y_test = clean_data(scaled=1)
# Create logistic regresion
logistic = LogisticRegression(solver='liblinear',
max_iter=150,
random_state=42)
# Create regularization penalty space
penalty = ['l1', 'l2']
# Create regularization hyperparameter space
# First run in logspace found C to be ~10
C = np.linspace(8, 12, 30)
# Create hyperparameter options
hyperparameters = dict(C=C, penalty=penalty)
# Create grid search using 5 split stratisfied shuffle split cross validation
cv = StratifiedShuffleSplit(n_splits=5, test_size=0.2, random_state=42)
clf = GridSearchCV(logistic,
hyperparameters,
cv=cv,
verbose=0,
scoring='precision',
n_jobs=-1)
# Fit grid search
best_model = clf.fit(X_train_scaled, y_train)
# View best hyperparameters
print('Best Penalty:', best_model.best_estimator_.get_params()['penalty'])
print('Best C:', best_model.best_estimator_.get_params()['C'])
print(f'Best training score {best_model.best_score_}')
# Make a prediction on entire training set
y_pred = best_model.best_estimator_.predict(X_test_scaled)
# Classification report showing precision,
report = classification_report(y_true=y_test, y_pred=y_pred)
print(report)
# Display confusion matrix
CM = confusion_matrix(y_true=y_test, y_pred=y_pred, labels=[0, 1])
cm(cm=CM,
target_names=['Burpee no jump', 'Burpee'],
title='Logistric Regression CM')
# Plot a bar graph of the variables to get insight in importance
bar_coef(Model=best_model, X_train_scaled=X_test_scaled)
# Plot ROC curve
ROC(Model=best_model.best_estimator_, Y_test=y_test, X_test=X_test_scaled)
def run_model():
# Load data
X_train_scaled, y_train, X_test_scaled, y_test = clean_data(scaled=1)
# Create SVC
svc = SVC(max_iter=-1, probability=True)
# Create diff kernels
kernel = ['rbf', 'linear', 'poly', 'sigmoid']
# Degree for poly kernels
degree = np.arange(start=1, stop=5, step=1)
# Gamma for poly kernels
gamma = np.logspace(start=-15, stop=4, num=18, base=2)
# C penalty factor
C = np.logspace(start=-3, stop=16, num=18, base=2)
# Create hyperparameter options
hyperparameters = dict(kernel=kernel, degree=degree, gamma=gamma, C=C)
# Create grid search using 5 split stratisfied shuffle split cross validation
cv = StratifiedShuffleSplit(n_splits=10, test_size=0.2, random_state=42)
clf = GridSearchCV(svc,
hyperparameters,
cv=cv,
verbose=1,
scoring='precision',
n_jobs=-1)
# Fit grid search
best_model = clf.fit(X_train_scaled, y_train)
# View best hyperparameters
print(f'Best parameters {best_model.best_params_}')
print(f'Train score is {best_model.best_score_}')
# Best parameters {'C': 0.125, 'degree': 2, 'gamma': 0.07063223646433309, 'kernel': 'poly'}
# Make a prediction on entire training set
y_pred = best_model.best_estimator_.predict(X_test_scaled)
# Classification report
report = classification_report(y_true=y_test, y_pred=y_pred)
print(report)
# Confusion Matrix
CM = confusion_matrix(y_true=y_test, y_pred=y_pred, labels=[0, 1])
cm(cm=CM, target_names=['Burpee no jump', 'Burpee'], title='SVM CM')
# ROC
ROC(Model=best_model.best_estimator_, Y_test=y_test, X_test=X_test_scaled)
def run_model():
X_train_scaled, y_train, X_test_scaled, y_test = clean_data(scaled=1)
# Create knn
knn = KNeighborsClassifier()
# Create n neighbors hyperparameter space
n_neighbors = np.arange(start=1, stop=30, step=1)
# Create hyperparameter options
hyperparameters = dict(n_neighbors=n_neighbors)
# Create grid search using 5 split stratisfied shuffle split cross validation
cv = StratifiedShuffleSplit(n_splits=10, test_size=0.2, random_state=42)
clf = GridSearchCV(knn,
hyperparameters,
cv=cv,
verbose=0,
scoring='precision',
n_jobs=-1)
# Fit grid search
best_model = clf.fit(X_train_scaled, y_train)
# View best hyperparameters
print('Best N:', best_model.best_estimator_.get_params()['n_neighbors'])
print(f'Train score is {best_model.best_score_}')
cv_scores = best_model.cv_results_['mean_test_score']
plt.plot(n_neighbors, cv_scores)
plt.xlabel('K'), plt.ylabel('Mean Test Score - Precision')
# Make a prediction on entire training set
y_pred = best_model.best_estimator_.predict(X_test_scaled)
# Classification Report
report = classification_report(y_true=y_test, y_pred=y_pred)
print(report)
# Confusion Matrix
CM = confusion_matrix(y_true=y_test, y_pred=y_pred, labels=[0, 1])
cm(cm=CM, target_names=['Burpee no jump', 'Burpee'], title='KNN CM')
# ROPC
ROC(Model=best_model.best_estimator_, Y_test=y_test, X_test=X_test_scaled)
def run_model():
# Load data
X_train_scaled, y_train, X_test_scaled, y_test = clean_data(scaled=1)
# Create knn
rf = RandomForestClassifier(random_state=42)
# Hyperparameter space for RF
# Number of trees in random forest
n_estimators = n_estimators = np.arange(start=1, stop=30, step=1)
# Number of features to consider at every split
max_features = ['auto', 'sqrt']
# Maximum number of levels in tree
max_depth = [int(x) for x in np.linspace(10, 110, num=11)]
max_depth.append(None)
# Minimum number of samples required to split a node
min_samples_split = [2, 5, 10]
# Minimum number of samples required at each leaf node
min_samples_leaf = [1, 2, 4]
# Method of selecting samples for training each tree
bootstrap = [True, False]
# Create the random grid
random_grid = {
'n_estimators': n_estimators,
'max_features': max_features,
'max_depth': max_depth,
'min_samples_split': min_samples_split,
'min_samples_leaf': min_samples_leaf,
'bootstrap': bootstrap
}
# Create grid search using 5 split stratisfied shuffle split cross validation
cv = StratifiedShuffleSplit(n_splits=5, test_size=0.2, random_state=42)
clf = rf_random = RandomizedSearchCV(estimator=rf,
param_distributions=random_grid,
n_iter=100,
cv=3,
verbose=2,
random_state=42,
n_jobs=-1,
scoring='precision')
# Fit grid search
best_model = clf.fit(X_train_scaled, y_train)
# View best hyperparameters
print(f'Best parameters {best_model.best_params_}')
print(f'Train score is {best_model.best_score_}')
# Best parameters {'n_estimators': 17, 'min_samples_split': 10, 'min_samples_leaf': 1, 'max_features': 'sqrt', 'max_depth': 80, 'bootstrap': False}
# Train score is 0.9477959255356516
# Make a prediction on entire training set
y_pred = best_model.best_estimator_.predict(X_test_scaled)
report = classification_report(y_true=y_test, y_pred=y_pred)
print(report)
# Confusion Matrix
CM = confusion_matrix(y_true=y_test, y_pred=y_pred, labels=[0, 1])
cm(cm=CM,
target_names=['Burpee no jump', 'Burpee'],
title='Random Forest CM')
# ROC
ROC(Model=best_model.best_estimator_, Y_test=y_test, X_test=X_test_scaled)