svm.py

# Import necessary packages
import os
import numpy as np
import pandas as pd
from sklearn.svm import SVC
from sklearn.model_selection import GridSearchCV
from sklearn import metrics

def load_data():
    """ load data from CSV formats """
    path = 'data/real_signal'
    train = os.path.join(path, 'train_data_svm.txt')
    test = os.path.join(path, 'test_data_svm.txt')
    train_class = os.path.join(path, 'train_class_svm.txt')
    test_class = os.path.join(path, 'test_class_svm2.txt')

    Xtr = np.array(pd.read_csv(train, header=None))
    Xte = np.array(pd.read_csv(test, header=None))
    ytr = np.array(pd.read_csv(train_class, header=None))
    yte = np.array(pd.read_csv(test_class, header=None))

    return Xtr, Xte, ytr.reshape(-1), yte.reshape(-1)

def normal_svm():
    # candidates of hyperparameters for grid search
    param_grid = {'C': [2**i for i in range(5, 10)],
                  'gamma': [2**i for i in range(5, 10)]}

    Xtr, Xte, ytr, yte = load_data() # load data
    # declare a model
    model = GridSearchCV(SVC(), param_grid=param_grid, n_jobs=-1, cv=10)
    model.fit(Xtr, ytr) # fit the training data

    # calculate decision function values and prediction
    score = model.decision_function(Xte)
    ypred = model.predict(Xte)
    
    # save the best hyperparameter and several performance metrics in CSV
    result = pd.DataFrame(columns=['C', 'gamma', 'accuracy', 'AUC', 'precision',
                                   'recall', 'F1', 'Nsv'])

    C = model.best_params_['C']
    gamma = model.best_params_['gamma']
    acc = metrics.accuracy_score(yte, ypred)
    auc = metrics.roc_auc_score(yte, score)
    prec = metrics.precision_score(yte, ypred)
    rec = metrics.recall_score(yte, ypred)
    f1 = metrics.f1_score(yte, ypred)
    nsv = sum(model.best_estimator_.n_support_)

    result.loc['Normal'] = [C, gamma, acc, auc, prec, rec, f1, nsv]
    print(result)
    result.to_csv('result.csv')

    # save the detailed results of cross validation
    cv_result = pd.DataFrame(model.cv_results_)
    cv_result.to_csv('cv_results.csv')

if __name__ == "__main__":
    normal_svm()