def k_nearest_neighbour(k):
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
correct = 0
total_per_digit = [0] * 10
correct_per_digit = [0] * 10
errors = open('./errors.txt', 'w')
for test in testcases:
total_per_digit[int(test[0])] += 1
classification = algorithms.k_nearest_neighbour(k, test, trainingset)
if classification == test[0]:
correct_per_digit[int(test[0])] += 1
correct += 1
else:
errors.write('Misclassified ' + test[0] + ' as ' + classification +
'\n')
print 'The overall recognition rate is: ' + str(
correct / float(len(testcases)))
print 'Broken down by digit:'
for i, num in enumerate(correct_per_digit):
print 'For digit ' + str(i) + ': ' + str(
num / float(total_per_digit[i]))
def prepare_training_set():
""" Loads the training set and prepares the data """
raw_training_data = read_training_data()
training_set = []
training_solution_set = []
# iterate through all training example
for example in raw_training_data:
image = example[0]
solution = example[1]
characters = extract_characters(image)
for i in range(len(characters)):
# only add classes we have not yet had an example before. A little hacky, but minimizes model size
if not solution[i] in training_solution_set:
training_set.append(characters[i].ravel())
training_solution_set.append(solution[i])
clf = SVC(C=100, gamma=0.0001)
clf.fit(training_set, training_solution_set)
joblib.dump(clf, 'model.pkl', compress=9)
def k_nearest_neighbour(k):
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
correct = 0
total_per_digit = [0] * 10
correct_per_digit = [0] * 10
errors = open('./errors.txt', 'w')
for test in testcases:
total_per_digit[int(test[0])] += 1
classification = algorithms.k_nearest_neighbour(k, test, trainingset)
if classification == test[0]:
correct_per_digit[int(test[0])] += 1
correct += 1
else:
errors.write('Misclassified ' + test[0] + ' as ' + classification + '\n')
print 'The overall recognition rate is: ' + str(correct / float(len(testcases)))
print 'Broken down by digit:'
for i, num in enumerate(correct_per_digit):
print 'For digit ' + str(i) + ': ' + str(num / float(total_per_digit[i]))
def linear_regression():
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
algorithms.linear(testcases, trainingset)
def rnd_for():
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
algorithms.random_forest(testcases, trainingset)
def naive_bayes():
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
algorithms.naive_bayes(testcases, trainingset)
def support_vector_machine():
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
algorithms.support_vector_machine(testcases, trainingset)
def linear_regression():
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
algorithms.linear(testcases, trainingset)
def rnd_for():
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
algorithms.random_forest(testcases, trainingset)
def naive_bayes():
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
algorithms.naive_bayes(testcases, trainingset)
def support_vector_machine():
_, trainingset = readdata.read_training_data()
_, testcases = readdata.read_test_data()
algorithms.support_vector_machine(testcases, trainingset)