def test_model(notes, predictions, clf):
print("\tGetting note vectors...")
note_vectors = [note_to_vec(note) for note in notes]
print("\tSplitting train test...")
X_train, X_test, y_train, y_test = train_test_split(note_vectors,
predictions,
test_size=0.2)
print("\tFitting...")
clf.fit(X_train, y_train)
print("\tPredicting...")
generated_predictions = clf.predict(X_test)
return accuracy_score(y_test, generated_predictions)
def visualize_pca_by_lang(notes):
_prepare_plots()
pca = __get_pca_model(notes)
langs = [note.get_lang() for note in notes]
langs = sorted(set(langs))
for lang in langs:
lang_notes = [note for note in notes if note.get_lang() == lang]
lang_notes_vectors = [note_to_vec(note) for note in lang_notes]
pca_lang_notes = pca.transform(lang_notes_vectors)
plt.scatter(pca_lang_notes[:, 0], pca_lang_notes[:, 1], alpha=0.7, label=lang)
plt.legend()
_save_plot('lang')
def visualize_pca_by_growth(notes):
_prepare_plots()
pca = __get_pca_model(notes)
growths = [note.growth for note in notes]
growths = sorted(set(growths))
colors = iter(['#e74c3c', '#2ecc71'])
for growth in growths:
growth_notes = [note for note in notes if note.growth == growth]
growth_notes_vectors = [note_to_vec(note) for note in growth_notes]
pca_growth_notes = pca.transform(growth_notes_vectors)
plt.scatter(pca_growth_notes[:, 0],
pca_growth_notes[:, 1], label=growth, alpha=0.7, color=next(colors))
plt.legend()
_save_plot('growth')
def visualize_pca_by_currency(notes):
_prepare_plots()
pca = __get_pca_model(notes)
currencies = [note.currency for note in notes]
currencies = sorted(set(currencies))
colors = iter(sns.color_palette("hls", len(currencies)).as_hex())
for currency in currencies:
currency_notes = [note for note in notes if note.currency == currency]
currency_notes_vectors = [note_to_vec(note) for note in currency_notes]
pca_currency_notes = pca.transform(currency_notes_vectors)
plt.scatter(pca_currency_notes[:, 0],
pca_currency_notes[:, 1], label=currency, alpha=0.7, color=next(colors))
plt.legend()
_save_plot('currency')
def __get_pca_model(notes):
pca = PCA(n_components=2)
notes_vectors = [note_to_vec(note) for note in notes]
pca.fit(notes_vectors)
return pca