def my_LogisticRegression(X_train,Y_train,X_test,Y_test):
C = [0.01, 0.1, 1, 10]
i = 1
for param in C:
log_reg = linear_model.LogisticRegression(C=param, solver='lbfgs')
log_reg.fit(X_train, Y_train)
Y_predict_test = log_reg.predict(X_test)
y_predict_prob = log_reg.predict_proba(X_test)[:, 1]
false_positive_rate, true_positive_rate, thresholds = metrics.roc_curve(Y_test, y_predict_prob)
plt.subplot(2, 2, i)
plt.tight_layout()
plt.plot(false_positive_rate, true_positive_rate, i)
plt.xlim([-0.2, 1.2])
plt.ylim([-0.2, 1.2])
plt.title('ROC curve C = {}'.format(param))
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.grid(True)
# Use score method to get accuracy of model
score = log_reg.score(X_test, Y_test)
print("+ C = %f => Test accuracy: %f" % (param, score))
print_performance_metrics(Y_predict_test,Y_test)
i = i + 1
print("* Save Logistic Regression classification result into image/logistic_regression.png")
plt.savefig('image/logistic_regression.png')
plt.show()
# define a function that accepts a threshold and prints sensitivity and specificity
def evaluate_threshold(threshold):
print('Sensitivity:', true_positive_rate[thresholds > threshold][-1])
print('Specificity:', 1 - false_positive_rate[thresholds > threshold][-1])
evaluate_threshold(0.9)
def my_NonLinearSVM(X_train, Y_train, X_test, Y_test):
C = [0.01, 0.1, 0.2, 0.5, 0.8, 1, 5, 10, 20, 50]
NL_SVM_train_Accuracy = []
NL_SVM_test_Accuracy = []
for param in C:
clf = SVC(C=param, kernel='rbf', gamma='auto')
clf.fit(X_train, Y_train)
Y_predict_NL_SVMTrain = clf.predict(X_train)
Y_predict_NL_SVM_Test = clf.predict(X_test)
NL_SVM_train_Accuracy.append(
accuracy_score(Y_train, Y_predict_NL_SVMTrain))
NL_SVM_test_Accuracy.append(
accuracy_score(Y_test, Y_predict_NL_SVM_Test))
print(" * C = %f => Train accuracy = %f and Test accuracy = %f " %
(param, accuracy_score(Y_train, Y_predict_NL_SVMTrain),
accuracy_score(Y_test, Y_predict_NL_SVM_Test)))
print_performance_metrics(Y_predict_NL_SVM_Test, Y_test)
plt.plot(C, NL_SVM_train_Accuracy, 'ro-')
plt.plot(C, NL_SVM_test_Accuracy, 'bv--')
plt.legend(['Training Accuracy', 'Test Accuracy'])
plt.xlabel('C')
plt.xscale('log')
plt.ylabel('Accuracy')
plt.title('Nonlinear Support Vector Machine')
print(
"* Save Non Linear Support Vector Machine classification result into image/non_linear_svm.png"
)
plt.savefig('image/non_linear_svm.png')
plt.show()
def my_SVM(X_train, Y_train, X_test, Y_test):
# Support Vector Machine Classifier
plt.clf()
C = [0.01, 0.1, 0.2, 0.5, 0.8, 1, 5, 10, 20, 50]
SVM_train_accuracy = []
SVM_test_accuracy = []
for param in C:
clf = SVC(C=param, kernel='linear')
clf.fit(X_train, Y_train)
Y_predict_SVM_train = clf.predict(X_train)
Y_predict_SVM_test = clf.predict(X_test)
SVM_train_accuracy.append(accuracy_score(Y_train, Y_predict_SVM_train))
SVM_test_accuracy.append(accuracy_score(Y_test, Y_predict_SVM_test))
print("+ C = %f => Train accuracy = %f and Test accuracy = %f " %
(param, accuracy_score(Y_train, Y_predict_SVM_train),
accuracy_score(Y_test, Y_predict_SVM_test)))
print_performance_metrics(Y_predict_SVM_test, Y_test)
plt.plot(C, SVM_train_accuracy, 'ro-')
plt.plot(C, SVM_test_accuracy, 'bv--')
plt.legend(['Training Accuracy', 'Test Accuracy'])
plt.xlabel('C')
plt.xscale('log')
plt.ylabel('Accuracy')
plt.title('Support Vector Machine')
print(
"* Save Support Vector Machine classification result into image/svm.png"
)
plt.savefig('image/svm.png')
plt.show()
def my_KNeighborsClassifier(X_train_std,X_test,y_train,y_test):
acc_list_train=[]
acc_list_test=[]
n=[i for i in range(1,51)]
for i in range(1,51):
knn = KNeighborsClassifier(n_neighbors=i)
knn.fit(X_train_std, y_train)
acc_list_train.append(accuracy_score(y_train,knn.predict(X_train_std)))
acc_list_test.append(accuracy_score(y_test,knn.predict(X_test)))
fig,axes_knn=plt.subplots(1,1)
plt.plot(n,acc_list_train,color='Red',label='Training accuracy')
plt.plot(n,acc_list_test,color='Blue',label='Testing accuracy')
plt.ylabel('Accuracy Score')
plt.xlabel('n (Nearest Neighbors)')
plt.title('KNN')
plt.legend(loc='upper right')
print("* Save KNN classification result into image/knn.png")
plt.savefig('image/knn.png')
max_indexes_test=[i+1 for i,value in enumerate(acc_list_test) if value == max(acc_list_test)]
print("* Best K is "+ str(max_indexes_test[0]) )
clf = KNeighborsClassifier(n_neighbors=max_indexes_test[0], metric='euclidean', p=2)
clf.fit(X_train_std, y_train)
Y_predTrain = clf.predict(X_train_std)
Y_predTest_best = clf.predict(X_test)
print("* Training accuracy : ",accuracy_score(y_train, Y_predTrain))
print("* Testing accuracy : ",accuracy_score(y_test, Y_predTest_best))
print_performance_metrics(Y_predTest_best,y_test)
def my_GaussianNB(X_train,Y_train,X_test,Y_test):
gnb = GaussianNB()
gnb.fit(X_train, Y_train)
Y_predTrain = gnb.predict(X_train)
y_predTest = gnb.predict(X_test)
print("* Training accuracy ",accuracy_score(Y_train, Y_predTrain))
print("* Testing accuracy",accuracy_score(Y_test, y_predTest))
print_performance_metrics(y_predTest,Y_test)
def my_RandomForestClassifier(X_train, X_test, Y_train, Y_test,
numBaseClassifiers, train_Acc, test_Acc):
clf = ensemble.RandomForestClassifier(n_estimators=numBaseClassifiers)
clf.fit(X_train, Y_train)
Y_predict_train_EM = clf.predict(X_train)
Y_predict_test_EM = clf.predict(X_test)
train_Acc.append(accuracy_score(Y_train, Y_predict_train_EM))
test_Acc.append(accuracy_score(Y_test, Y_predict_test_EM))
print("* Train accuracy = %f and Test accuracy = %f " %
(accuracy_score(Y_train, Y_predict_train_EM),
accuracy_score(Y_test, Y_predict_test_EM)))
print_performance_metrics(Y_predict_test_EM, Y_test)
def my_Bagging(X_train, X_test, Y_train, Y_test, numBaseClassifiers,
max_depth_EM, train_Acc, test_Acc):
clf = ensemble.BaggingClassifier(
DecisionTreeClassifier(max_depth=max_depth_EM),
n_estimators=numBaseClassifiers)
clf.fit(X_train, Y_train)
Y_predict_train_EM = clf.predict(X_train)
Y_predict_test_EM = clf.predict(X_test)
train_Acc.append(accuracy_score(Y_train, Y_predict_train_EM))
test_Acc.append(accuracy_score(Y_test, Y_predict_test_EM))
print("* Train accuracy = %f and Test accuracy = %f " %
(accuracy_score(Y_train, Y_predict_train_EM),
accuracy_score(Y_test, Y_predict_test_EM)))
print_performance_metrics(Y_predict_test_EM, Y_test)
def my_DecisionTree(X_train, X_test, y_train, y_test):
print('* Use Entropy index for impurity measure :')
accuracy = np.empty(2, dtype=float)
max_depths = [2, 3]
i = 0
for max_depth in max_depths:
clf = tree.DecisionTreeClassifier(criterion='entropy',
max_depth=max_depth)
clf = clf.fit(X_train, y_train)
# create graph tree with class names
dot_data = tree.export_graphviz(clf,
feature_names=X_train.columns,
class_names=['1', '0'],
filled=True,
out_file=None)
graph = pydotplus.graph_from_dot_data(dot_data)
graph.write_png('image/tree_entropy_%d.png' % (max_depth))
predY = clf.predict(X_test)
accuracy[i] = accuracy_score(y_test, predY)
print('+ Entropy: Max depth %d , Accuracy on test data is %.2f' %
(max_depth, (accuracy[i])))
print_performance_metrics(predY, y_test)
i += 1
# plt.clf()
fig, ax = plt.subplots(nrows=1, ncols=2)
ax[0].set_prop_cycle(color=['red'])
ax[0].set_ylim([0.8, 1.1])
ax[0].plot(max_depths, accuracy)
ax[0].legend(['accuracy-Entropy'], loc='upper left')
# Using Gini index as impurity measure, fit decision trees of different maximum depths [2, 3, 4, 5,
# 6, 7, 8, 9, 10, 15, 20, 25] to the training set
i = 0
print('* Use Gini index for impurity measure :')
for max_depth in [2, 3]:
clf = tree.DecisionTreeClassifier(criterion='gini',
max_depth=max_depth)
clf = clf.fit(X_train, y_train)
dot_data = tree.export_graphviz(clf,
feature_names=X_train.columns,
class_names=['1', '0'],
filled=True,
out_file=None)
graph = pydotplus.graph_from_dot_data(dot_data)
graph.write_png('image/tree_gini_%d.png' % (max_depth))
predY = clf.predict(X_test)
accuracy[i] = accuracy_score(y_test, predY)
print('+ Gini: Max depth %d , Accuracy on test data is %.2f' %
(max_depth, (accuracy_score(y_test, predY))))
print_performance_metrics(predY, y_test)
i += 1
ax[1].set_prop_cycle(color=['green'])
ax[1].set_ylim([0.8, 1.1])
ax[1].plot(max_depths, accuracy)
ax[1].legend(['accuracy-Gini'], loc='upper left')
plt.title('Decision Tree')
print(
"* Save Decision Tree classification result into image/decision_tree.png"
)
plt.savefig('image/decision_tree.png')
plt.show()