ax.tick_params(labelsize=14)
plt.savefig(f'../model/tuning/{NAME}-NB.png', dpi=300)
if __name__ == '__main__':
base_params = {'alpha': 1.0, 'fit_prior': True, 'class_prior': None}
params_NB = dict(base_params)
param_grid_ = {'alpha': [0.001, 0.01, 0.1, 1, 10, 100]}
features = ["bow", "n-gram", "tf-idf", "n-gram-tf-idf"]
results = []
NAME = ":".join(features)
for name in features:
x = Runner.load_x_train(name)
y = Runner.load_y_train()
model = ModelMultinomialNB(name, **dict(params_NB))
search = GridSearchCV(model,
cv=6,
param_grid=param_grid_,
return_train_score=True,
verbose=10,
refit=True)
search.fit(x, y)
results.append((search, name))
logger.info(
f'{name} - bestscore : {search.best_score_} - result :{search.cv_results_["mean_test_score"]}'
)
res = pd.DataFrame.from_dict(
{
scaler.fit(X)
standard_X = scaler.transform(X)
# 12次元 -> 6次元 へ落とす.
dim = 6
params = {
'n_components': dim,
'random_state': 71,
}
# 主成分分析
clf = PCA(**params)
clf.fit(standard_X)
pca = clf.transform(standard_X)
pca = pca[:100]
labels = Runner.load_y_train()
labels = labels[:100]
happy = np.empty((0, dim))
angry = np.empty((0, dim))
sad = np.empty((0, dim))
#0=sad, 1=angry, 2=happy, 3=normal
for label, x in zip(labels, pca):
if label == 0:
sad = np.vstack((sad, x))
elif label == 1:
angry = np.vstack((angry, x))
elif label == 2:
happy = np.vstack((happy, x))
