svm.py

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
# https://gist.github.com/sithu/4722649d23c83440f2067ed429fa434b
import utils
from sklearn.svm import SVC

# Loading the linear dataset
# linear.csv: https://gist.github.com/sithu/701d1182d63b01e740bb244d8059ceb1
linear_data = pd.read_csv('linear.csv')
features = np.array(linear_data[['x_1', 'x_2']])
labels = np.array(linear_data['y'])
utils.plot_points(features, labels)

# Find accuracy
svm_linear = SVC(kernel='linear')
svm_linear.fit(features, labels)
print("Accuracy:", svm_linear.score(features, labels))
utils.plot_model(features, labels, svm_linear)

# Set C - hyperparameter
# C = 0.01 and C = 100: Which C gives better accuracy?
svm_c_001 = SVC(kernel='linear', C=0.01)
svm_c_001.fit(features, labels)
print("C = 0.01")
print("Accuracy:", svm_c_001.score(features, labels))
utils.plot_model(features, labels, svm_c_001)