elif l == 1 and h == -1:
tp += 1
elif l == -1 and h == 1:
fn += 1
else:
tn += 1
print '-----------------------------'
print '\tConfusion Matrix'
print '-----------------------------'
print '\t\tPredicted'
print '\tActual\tNO\tYES'
print '-----------------------------'
print '\tNO\t', tn, '\t', fp
print '-----------------------------'
print '\tYES\t', fn, '\t', tp
print '-----------------------------'
clf = svm_basic(5000)
dataArr, labelArr = loadDataSet('linearly_separable.csv')
size = int(len(dataArr) * 0.8)
train_data = dataArr[:size]
train_label = labelArr[:size]
test_data = dataArr[size:]
test_label = labelArr[size:]
weightArray = clf.fit(train_data, train_label)
plot_fit(weightArray, train_data, train_label)
hyp = clf.predict(test_data)
print('Accuracy: ', accuracy(test_label, hyp))
print_confusion_matrix(test_label, hyp)
import pandas as pd
import numpy as np
from svm_basic import svm_basic
data = pd.read_csv('linearly_separable.csv', delimiter=",", header=None)
Xin = data.iloc[:, :2]
Yin = data.iloc[:, 2]
dataArr = Xin.values.tolist()
labelArr = Yin.values.tolist()
splitpoint = int(len(labelArr) * 0.5)
train_x = dataArr[:splitpoint]
train_y = labelArr[:splitpoint]
test_x = dataArr[splitpoint:]
test_y = labelArr[splitpoint:]
svm = svm_basic(1, 0.001, 50)
svm.fit(train_x, train_y)
w0 = svm.weights.item(0)
w1 = svm.weights.item(1)
b = svm.b.item(0)
print("parameters: ")
print(w0, w1, b)
svm.predict(train_x, train_y)
svm.predict(test_x, train_y)
n = Xin.shape[0]
xcord1 = []
ycord1 = []
y = a * x - b[0,0]/weights[1]
ax.fill_between(x, y, y2=min_y - 1, color='red', alpha=0.5)
ax.fill_between(x, y, y2=max_y + 1, color='blue', alpha=0.5)
plt.tight_layout()
ax.axis([min_x - 1, max_x + 1, min_y - 1, max_y + 1])
ax.plot(x, y)
plt.xlabel('x1')
plt.ylabel('x2')
plt.savefig(name)
if __name__ == '__main__':
svm_classifier = svm_basic(maxIter=100)
X, Y = get_data_and_labels('linearly_separable.csv')
plot_data(X, Y, name='Linearly Seperable Data')
print('Saving initial scatter plot of data as Linearly Seperable Data.png')
print('Splitting data')
train_x, test_x, train_y, test_y = train_test_split(X, Y, test_size=0.2)
weights, b = svm_classifier.fit(train_x, train_y)
hyp = svm_classifier.predict(test_x)
print('Accuracy: {0:.4f}'.format(accuracy_score(test_y.values, hyp)))
print(confusion_matrix(test_y, hyp))
plot_data(X, Y, weights, b, plot_line=True, name='Linearly Separated Data')
print('Saving final result as Linearly Separated Data.png')
dataMat = []
labelMat = []
fr = open(fileName)
for line in fr.readlines():
lineArr = line.strip().split(',')
dataMat.append([float(lineArr[0]), float(lineArr[1])])
labelMat.append(float(lineArr[2]))
return dataMat, labelMat
X, Y = loadDataSet('linearly_separable.csv')
plt.scatter([i[0] for i in X], [i[1] for i in X], c=Y, alpha=0.5)
plt.savefig("UsingLabels.png")
clf = svm_basic()
b, weights = clf.fit(X, Y)
hypotheses = clf.predict(np.mat(X))
labels = []
x = np.arange(min([i[0] for i in X]), max([i[0] for i in X]), 0.5)
list_x = x.tolist()
list_x = np.append(list_x, [0, -float(b) / float(weights[1])])
x = np.array(list_x)
y = []
for i in range(0, len(x)):
y.append((-float(b) - (float(weights[1]) * list_x[i])) / float(weights[0]))
for i in range(0, len(hypotheses)):
if hypotheses[i] >= 0:
tp += 1.0
elif l == 1 and h == 0:
tn += 1.0
elif l == 0 and h == 0:
fn += 1.0
else:
tn += 1
print('-----------------------------')
print('\tConfusion Matrix')
print('-----------------------------')
print('\t\tPredicted')
print('\tActual\tNO\tYES')
print('-----------------------------')
print('\tNO\t', tn, '\t', fp)
print('-----------------------------')
print('\tYES\t', fn, '\t', tp)
print('-----------------------------')
dataArr, labelArr = get_data('linearly_separable.csv')
test = svm_basic(mat(dataArr), mat(labelArr).transpose(), 1, 0.001)
b, alphas = test.smoPK(50)
w = test.fit()
w = asmatrix(w)
hypotheses = test.predict(dataArr)
print("b =", b)
print("weights =\n", w)
plot_fit(w, dataArr, labelArr, squeeze(asarray(b)))
print('Accuracy:', accuracy(labelArr, hypotheses))
print_confusion_matrix(labelArr, hypotheses)