class TestLogisticRegression(unittest.TestCase):
samples, labels = dataparser.parse_dataset("/Users/droy/Downloads/dataset.csv")
split_index = int(len(samples) * 0.75)
train_dataset = samples[:split_index]
train_labels = labels[:split_index]
test_dataset = samples[split_index:]
test_labels = labels[split_index:]
test_data = samples[0]
def setUp(self):
self.logreg = LogisticRegressor(self.train_dataset, self.train_labels)
def test_gradient(self):
"""Just testing that it doesn't throw any errors"""
g = self.logreg.get_gradient()
print g(ones(len(self.samples[0]) + 1))
def test_gradient_descent(self):
scale = Scale.scale_from_data(self.test_dataset)
# self.logreg = LogisticRegressor(self.train_dataset, self.train_labels, scale=scale)
self.logreg = LogisticRegressor(self.train_dataset, self.train_labels)
gr = GradientDescentRunner(self.logreg.get_gradient(), len(self.samples[0]) + 1,
self.logreg.get_error_function(), alpha=1e-8, max_iter=300)
weights = gr.run()
self.logreg.weights = weights
mismatches = 0
predictions = [self.logreg.predict(test_data) for test_data in self.test_dataset]
print "All predictions: ", predictions
for test_data, label in zip(self.test_dataset, self.test_labels):
prediction = self.logreg.predict(test_data)
if prediction != label:
mismatches +=1
# print "Mismatch! Predicted ", prediction, ", True ", label
total = len(self.test_dataset)
print "total data: ", total
print "total mismatch: ", mismatches
print "percentage success: ", (100 - float(mismatches) / float(total) * 100), "%"
def test_sckikit_performance(self):
from sklearn import linear_model
clf = linear_model.LogisticRegression()
clf.fit(self.train_dataset, self.train_labels)
predictions = clf.predict(self.test_dataset)
print "Scikit predictions: ", predictions
mismatches = 0
true_positives = 0
true_negatives = 0
false_positives = 0
false_negatives = 0
for test_data, label in zip(self.test_dataset, self.test_labels):
prediction = clf.predict(test_data)
if prediction != label:
mismatches += 1
if prediction == 1 and label == 0:
false_positives += 1
if prediction == 0 and label == 1:
false_negatives += 1
if prediction == 1 and label == 1:
true_positives += 1
if prediction == 0 and label == 0:
true_negatives += 1
# print "Mismatch! Predicted ", prediction, ", True ", label
total = len(self.test_dataset)
print "True positives: ", true_positives
print "True negatives: ", true_negatives
print "False negatives: ", false_negatives
print "False positives: ", false_positives
print "F1 metric: ", 2 * float(true_positives) \
/ float(2 * true_positives + false_positives + false_negatives)
print "MCC: ", float(true_positives * true_negatives - false_positives * false_negatives) \
/ math.sqrt((true_positives + false_positives) * (true_positives + false_negatives) \
* (true_negatives + false_positives) * (true_negatives + false_negatives))
print "total data: ", total
print "total mismatch: ", mismatches
print "percentage success: ", (100 - float(mismatches) / float(total) * 100), "%"
def test_lets_just_look_at_the_outputs(self):
gr = GradientDescentRunner(self.logreg.get_gradient(), len(self.samples[0]) + 1,
self.logreg.get_error_function(), alpha=1e-8, max_iter=300)
_, weights = gr.run_once()
self.logreg.weights = weights
predictions = [self.logreg.get_probability(d) for d in self.train_dataset]
import pprint
pprint.pprint(zip(predictions, self.train_labels))
class TestLogisticRegression(unittest.TestCase):
samples, labels = dataparser.parse_dataset(
"/Users/droy/Downloads/dataset.csv")
split_index = int(len(samples) * 0.75)
train_dataset = samples[:split_index]
train_labels = labels[:split_index]
test_dataset = samples[split_index:]
test_labels = labels[split_index:]
test_data = samples[0]
def setUp(self):
self.logreg = LogisticRegressor(self.train_dataset, self.train_labels)
def test_gradient(self):
"""Just testing that it doesn't throw any errors"""
g = self.logreg.get_gradient()
print g(ones(len(self.samples[0]) + 1))
def test_gradient_descent(self):
scale = Scale.scale_from_data(self.test_dataset)
# self.logreg = LogisticRegressor(self.train_dataset, self.train_labels, scale=scale)
self.logreg = LogisticRegressor(self.train_dataset, self.train_labels)
gr = GradientDescentRunner(self.logreg.get_gradient(),
len(self.samples[0]) + 1,
self.logreg.get_error_function(),
alpha=1e-8,
max_iter=300)
weights = gr.run()
self.logreg.weights = weights
mismatches = 0
predictions = [
self.logreg.predict(test_data) for test_data in self.test_dataset
]
print "All predictions: ", predictions
for test_data, label in zip(self.test_dataset, self.test_labels):
prediction = self.logreg.predict(test_data)
if prediction != label:
mismatches += 1
# print "Mismatch! Predicted ", prediction, ", True ", label
total = len(self.test_dataset)
print "total data: ", total
print "total mismatch: ", mismatches
print "percentage success: ", (
100 - float(mismatches) / float(total) * 100), "%"
def test_sckikit_performance(self):
from sklearn import linear_model
clf = linear_model.LogisticRegression()
clf.fit(self.train_dataset, self.train_labels)
predictions = clf.predict(self.test_dataset)
print "Scikit predictions: ", predictions
mismatches = 0
true_positives = 0
true_negatives = 0
false_positives = 0
false_negatives = 0
for test_data, label in zip(self.test_dataset, self.test_labels):
prediction = clf.predict(test_data)
if prediction != label:
mismatches += 1
if prediction == 1 and label == 0:
false_positives += 1
if prediction == 0 and label == 1:
false_negatives += 1
if prediction == 1 and label == 1:
true_positives += 1
if prediction == 0 and label == 0:
true_negatives += 1
# print "Mismatch! Predicted ", prediction, ", True ", label
total = len(self.test_dataset)
print "True positives: ", true_positives
print "True negatives: ", true_negatives
print "False negatives: ", false_negatives
print "False positives: ", false_positives
print "F1 metric: ", 2 * float(true_positives) \
/ float(2 * true_positives + false_positives + false_negatives)
print "MCC: ", float(true_positives * true_negatives - false_positives * false_negatives) \
/ math.sqrt((true_positives + false_positives) * (true_positives + false_negatives) \
* (true_negatives + false_positives) * (true_negatives + false_negatives))
print "total data: ", total
print "total mismatch: ", mismatches
print "percentage success: ", (
100 - float(mismatches) / float(total) * 100), "%"
def test_lets_just_look_at_the_outputs(self):
gr = GradientDescentRunner(self.logreg.get_gradient(),
len(self.samples[0]) + 1,
self.logreg.get_error_function(),
alpha=1e-8,
max_iter=300)
_, weights = gr.run_once()
self.logreg.weights = weights
predictions = [
self.logreg.get_probability(d) for d in self.train_dataset
]
import pprint
pprint.pprint(zip(predictions, self.train_labels))