def spam():
train, target = load_spambase()
normalize_columns = [55, 56]
normalize(train, normalize_columns)
train = append_new_column(train, 1.0, 0)
# 10 fold cross validation
train_size = len(train)
k = 10
test_index_generator = cross_validation.k_fold_cross_validation(train_size, k)
fold = 0
train_accuracy = 0
test_accuracy = 0
train_mse = 0
test_mse = 0
for start, end in test_index_generator:
train_left = train[range(0, start)]
train_right = train[range(end, train_size)]
k_fold_train = np.vstack((train_left, train_right))
test = train[range(start, end)]
target_left = target[range(0, start)]
target_right = target[range(end, train_size)]
train_target = np.append(target_left, target_right)
test_target = target[range(start, end)]
cf = LinearRegression()
cf = cf.fit(k_fold_train, train_target)
print '=============Train Data Result============'
predict_train = cf.predict(k_fold_train)
cm = confusion_matrix(train_target, predict_train)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr, tpr)
train_accuracy += acc
print "mse: ", mse(predict_train, train_target), " rmse: ", rmse(predict_train, train_target), " mae: ", mae(
predict_train,
train_target)
train_mse += mse(predict_train, train_target)
print '=============Test Data Result============'
predict_test = cf.predict(test)
cm = confusion_matrix(test_target, predict_test)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr, tpr)
test_accuracy += acc
fold += 1
print "mse: ", mse(predict_test, test_target), " rmse: ", rmse(predict_test, test_target), " mae: ", mae(
predict_test,
test_target)
test_mse += mse(predict_test, test_target)
print "Average train acc: %f, average test acc: %f" % (train_accuracy / fold, test_accuracy / fold)
print "Average train mse: %f, average test mse: %f" % (train_mse / fold, test_mse / fold)
def gaussian_naive_bayes1():
train, target = load_spambase()
train, test, train_target, test_target = train_test_shuffle_split(train, target, len(train) / 10)
cf = GaussianNaiveBayes()
predicts = cf.fit(train, train_target).predict_class(test)
cm = confusion_matrix(test_target, predicts)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr, tpr)
def spam():
train, target = load_spambase()
normalize_columns = [55, 56]
normalize(train, normalize_columns)
train = append_new_column(train, 1.0, 0)
# 10 fold cross validation
train_size = len(train)
k = 10
test_index_generator = cross_validation.k_fold_cross_validation(
train_size, k)
fold = 0
train_accuracy = 0
test_accuracy = 0
train_mse = 0
test_mse = 0
for start, end in test_index_generator:
train_left = train[range(0, start)]
train_right = train[range(end, train_size)]
k_fold_train = np.vstack((train_left, train_right))
test = train[range(start, end)]
target_left = target[range(0, start)]
target_right = target[range(end, train_size)]
train_target = np.append(target_left, target_right)
test_target = target[range(start, end)]
cf = LinearRegression()
cf = cf.fit(k_fold_train, train_target)
print '=============Train Data Result============'
predict_train = cf.predict(k_fold_train)
cm = confusion_matrix(train_target, predict_train)
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
train_accuracy += acc
print "mse: ", mse(predict_train, train_target), " rmse: ", rmse(
predict_train,
train_target), " mae: ", mae(predict_train, train_target)
train_mse += mse(predict_train, train_target)
print '=============Test Data Result============'
predict_test = cf.predict(test)
cm = confusion_matrix(test_target, predict_test)
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
test_accuracy += acc
fold += 1
print "mse: ", mse(predict_test, test_target), " rmse: ", rmse(
predict_test, test_target), " mae: ", mae(predict_test,
test_target)
test_mse += mse(predict_test, test_target)
print "Average train mse: %f, average test mse: %f" % (
1.0 * train_mse / fold, 1.0 * test_mse / fold)
def decision_tree_all_data():
train, target = load_spambase()
cf = tree.DecisionTree()
cf = cf.fit(train, target, 5)
print_tree(cf.root)
predicts = cf.predict(train)
cm = confusion_matrix(target,predicts)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr, tpr)
def spambase(T=100):
train, target = load_spambase()
target = np.array(map(lambda v: -1.0 if v == 0 else 1.0, target))
train, test, train_target, test_target = train_test_shuffle_split(train, target, len(train) / 10)
boost = AdaBoost()
start = timeit.default_timer()
boost.boost(train, train_target, test, test_target, T)
stop = timeit.default_timer()
print "Total Run Time: %s secs" % (stop - start)
def spambase(T=100):
train, target = load_spambase()
target = np.array(map(lambda v: -1.0 if v == 0 else 1.0, target))
train, test, train_target, test_target = train_test_shuffle_split(
train, target,
len(train) / 10)
boost = AdaBoost()
start = timeit.default_timer()
boost.boost(train, train_target, test, test_target, T)
stop = timeit.default_timer()
print "Total Run Time: %s secs" % (stop - start)
def decision_tree_all_data():
train, target = load_spambase()
cf = tree.DecisionTree()
cf = cf.fit(train, target, 5)
print_tree(cf.root)
predicts = cf.predict(train)
cm = confusion_matrix(target, predicts)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (
cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr,
tpr)
def gaussian_naive_bayes1():
train, target = load_spambase()
train, test, train_target, test_target = train_test_shuffle_split(
train, target,
len(train) / 10)
cf = GaussianNaiveBayes()
predicts = cf.fit(train, train_target).predict_class(test)
cm = confusion_matrix(test_target, predicts)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (
cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr,
tpr)
def spam(step, loop, converge):
train, target = load_spambase()
train, test, train_target, test_target = cross_validation.train_test_shuffle_split(train, target, len(train) / 10)
scaler = normalize(train)
train = append_new_column(train, 1.0, 0)
scaler.scale_test(test)
test = append_new_column(test, 1.0, 0)
print '\n============== Logistic Regression - Stochastic Gradient Descending==============='
spam_logistic(train, test, train_target, test_target, step, loop, converge)
print '\n============== Linear Regression - Stochastic Gradient Descending ==============='
spam_linear(train, test, train_target, test_target, step, loop, converge)
print '\n============== Linear Regression - Normal Equation==============='
spam_normal_equation(train, test, train_target, test_target)
print '\n============== Decision Tree ===================================='
spam_decision_tree(train, test, train_target, test_target)
def naive_bayes(c, on_train=False):
train, target = load_spambase()
train, target = shuffle(train, target)
k = 10
train_size = len(train)
test_index_generator = k_fold_cross_validation(train_size, k)
fold = 1
overall_acc = 0
overall_error = 0
overall_auc = 0
overall_train_acc = 0
overall_train_error = 0
for start, end in test_index_generator:
k_fold_train = np.vstack((train[range(0, start)], train[range(end, train_size)]))
test = train[range(start, end)]
train_target = np.append(target[range(0, start)], target[range(end, train_size)])
test_target = target[range(start, end)]
cf = get_classifier(c)
raw_predicts = cf.fit(k_fold_train, train_target).predict(test)
predicts = cf.predict_class(raw_predicts)
if on_train:
raw_train_predicts = cf.predict(k_fold_train)
train_predicts = cf.predict_class(raw_train_predicts)
print '=============Fold %s Train==================' % fold
cm = confusion_matrix(train_target, train_predicts)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr, tpr)
overall_train_acc += acc
overall_train_error += er
print '=============Fold %s Test==================' % fold
cm = confusion_matrix(test_target, predicts)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr, tpr)
roc = ROC(test_target, raw_predicts, 0, "NaN")
recs = roc.create_roc_records()
roc.plot_roc_data(recs)
auc = roc.auc()
print 'AUC: %s' % auc
if fold == 1:
roc_data.append(roc.points)
overall_acc += acc
overall_error += er
overall_auc += auc
fold += 1
print '--------------- Result-------------------'
if on_train:
print 'Overall Accuracy: %s, Overall Error: %s' % (overall_train_acc/k, overall_train_error/k)
print 'Overall Accuracy: %s, Overall Error: %s, Overall AUC: %s\n' % (overall_acc/k, overall_error/k, overall_auc/k)
def naive_bayes(c, on_train=False):
train, target = load_spambase()
train, target = shuffle(train, target)
k = 10
train_size = len(train)
test_index_generator = k_fold_cross_validation(train_size, k)
fold = 1
overall_acc = 0
overall_error = 0
overall_auc = 0
overall_train_acc = 0
overall_train_error = 0
for start, end in test_index_generator:
k_fold_train = np.vstack(
(train[range(0, start)], train[range(end, train_size)]))
test = train[range(start, end)]
train_target = np.append(target[range(0, start)],
target[range(end, train_size)])
test_target = target[range(start, end)]
cf = get_classifier(c)
raw_predicts = cf.fit(k_fold_train, train_target).predict(test)
predicts = cf.predict_class(raw_predicts)
if on_train:
raw_train_predicts = cf.predict(k_fold_train)
train_predicts = cf.predict_class(raw_train_predicts)
print '=============Fold %s Train==================' % fold
cm = confusion_matrix(train_target, train_predicts)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (
cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc,
fpr, tpr)
overall_train_acc += acc
overall_train_error += er
print '=============Fold %s Test==================' % fold
cm = confusion_matrix(test_target, predicts)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (
cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr,
tpr)
roc = ROC(test_target, raw_predicts, 0, "NaN")
recs = roc.create_roc_records()
roc.plot_roc_data(recs)
auc = roc.auc()
print 'AUC: %s' % auc
if fold == 1:
roc_data.append(roc.points)
overall_acc += acc
overall_error += er
overall_auc += auc
fold += 1
print '--------------- Result-------------------'
if on_train:
print 'Overall Accuracy: %s, Overall Error: %s' % (
overall_train_acc / k, overall_train_error / k)
print 'Overall Accuracy: %s, Overall Error: %s, Overall AUC: %s\n' % (
overall_acc / k, overall_error / k, overall_auc / k)
def decision_tree():
train, target = load_spambase()
# 10 fold cross validation
train_size = len(train)
k = 10
test_index_generator = cross_validation.k_fold_cross_validation(
train_size, k)
fold = 0
train_accuracy = 0
test_accuracy = 0
train_mse = 0
test_mse = 0
for start, end in test_index_generator:
train_left = train[range(0, start)]
train_right = train[range(end, train_size)]
k_fold_train = np.vstack((train_left, train_right))
test = train[range(start, end)]
target_left = target[range(0, start)]
target_right = target[range(end, train_size)]
train_target = np.append(target_left, target_right)
test_target = target[range(start, end)]
cf = tree.DecisionTree()
cf = cf.fit(k_fold_train, train_target, 5)
print "=========Tree=============="
print_tree(cf.root)
print '=============Train Data Result============'
predict_train = cf.predict(k_fold_train)
cm = confusion_matrix(train_target, predict_train)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (
cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr,
tpr)
train_accuracy += acc
print "mse: ", mse(predict_train, train_target), " rmse: ", rmse(
predict_train,
train_target), " mae: ", mae(predict_train, train_target)
train_mse += mse(predict_train, train_target)
print '=============Test Data Result============'
predict_test = cf.predict(test)
cm = confusion_matrix(test_target, predict_test)
print "confusion matrix: TN: %s, FP: %s, FN: %s, TP: %s" % (
cm[0, 0], cm[0, 1], cm[1, 0], cm[1, 1])
er, acc, fpr, tpr = confusion_matrix_analysis(cm)
print 'Error rate: %f, accuracy: %f, FPR: %f, TPR: %f' % (er, acc, fpr,
tpr)
test_accuracy += acc
print "mse: ", mse(predict_test, test_target), " rmse: ", rmse(
predict_test, test_target), " mae: ", mae(predict_test,
test_target)
test_mse += mse(predict_test, test_target)
fold += 1
print "Average train acc: %f, average test acc: %f" % (
train_accuracy / fold, test_accuracy / fold)
print "Average train mse: %f, average test mse: %f" % (train_mse / fold,
test_mse / fold)