def spam_detection(random_state=0, fraction=1.0):
    vectorizer = CountVectorizer()
    h = gzip.open(get_path('ham.txt.gz'), 'rb')
    s = gzip.open(get_path('spam.txt.gz'), 'rb')
    hr = h.read()
    sr = s.read()
    ham = hr[0:int(fraction*len(hr))]
    spam = sr[0:int(fraction*len(sr))]
    hammat = vectorizer.fit_transform([ham])
    spammat = vectorizer.fit_transform([spam])
    #print(hammat.shape, spammat.shape)
    X = vectorizer.fit_transform([ham, spam])   #np.vstack([hammat, spammat])

    y = np.hstack([[0]*hammat.shape[0], [1]*spammat.shape[0]])   #[0]*hammat.shape[1] + [1]*spammat.shape[1]
    #print(X.shape, y.shape)


    X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=random_state, train_size=0.75)
    model = MultinomialNB()
    model.fit(X_train, y_train)
    y_fitted = model.predict(X_test)
    acc=metrics.accuracy_score(y_test, y_fitted)

    return acc, 0, 0