train_svm.py

# script to train VBL-VA001

from sklearn.svm import SVC
import matplotlib.pyplot as plt
from sklearn.neighbors import KNeighborsClassifier
from sklearn.model_selection import train_test_split
import numpy as np
import pandas as pd

# load data hasil ekstraksi fitur fft
x = pd.read_csv("data/feature_VBL-VA001.csv", header=None)

# load label
y = pd.read_csv("data/label_VBL-VA001.csv", header=None)

# make 1D array to avoid warning
y = pd.Series.ravel(y)


X_train, X_test, y_train, y_test = train_test_split(
    x, y, test_size=0.2, random_state=42, shuffle=True
)

print("Shape of Train Data : {}".format(X_train.shape))
print("Shape of Test Data : {}".format(X_test.shape))

# Setup arrays to store training and test accuracies
c_svm = np.arange(1, 100)
train_accuracy = np.empty(len(c_svm))
test_accuracy = np.empty(len(c_svm))

for i, k in enumerate(c_svm):
    # Setup a knn classifier with c_svm
    svm = SVC(C=k)
    # Fit the model
    svm.fit(X_train, y_train)
    # Compute accuracy on the training set
    train_accuracy[i] = svm.score(X_train, y_train)
    # Compute accuracy on the test set
    test_accuracy[i] = svm.score(X_test, y_test)

# Generate plot
# plt.title('Varying number of SVM')
plt.plot(c_svm, test_accuracy, label='Testing Accuracy')
plt.plot(c_svm, train_accuracy, label='Training accuracy')
plt.legend()
plt.xlabel('C')
plt.ylabel('Accuracy')
plt.show()
# np.savetxt('scm_c.txt', test_accuracy)
# plt.savefig('acc_svm.pdf')

# print optimal C and max test accuracy
print(f"Optimal C: {np.argmax(test_accuracy)}")
print(f"Max test accuracy: {max(test_accuracy)}")