def pred_label(n):
model = joblib.load('model.pkl')
vectorizer = joblib.load('vectorizer.pkl')
data_x, data_y = make_data_x_y()
text_vec = vectorizer.transform(data_x[:n])
for text, label in zip(text_vec, data_y):
print(f'{label}\t{model.predict(text)}\t{model.predict_proba(text)}')
def accuracy():
model = joblib.load('model.pkl')
vectorizer = joblib.load('vectorizer.pkl')
data_x, data_y = make_data_x_y()
data_x = vectorizer.transform(data_x)
data_y = np.array(data_y)
label_true, label_pred = data_y, model.predict(data_x)
print(classification_report(label_true, label_pred))
print(f'accuracy = {accuracy_score(label_true, label_pred)}')
def test(n):
model = joblib.load('model.pkl')
vectorizer = joblib.load('vectorizer.pkl')
data_x, data_y = make_data_x_y()
text_vec = vectorizer.transform(data_x[:n])
for vec, text in zip(text_vec, data_x):
print(f'入力文 : {text.strip()}')
print(f'予測結果 : {model.predict(vec)}')
print(f'予測確率 : {model.predict_proba(vec)}')
print()
def p_r_curve():
model = joblib.load('model.pkl')
vectorizer = joblib.load('vectorizer.pkl')
data_x, data_y = make_data_x_y()
data_x = vectorizer.transform(data_x)
data_y = np.array(data_y)
precision, recall, threshold = precision_recall_curve(
data_y,
model.predict_proba(data_x)[:, 1])
plt.plot(precision, recall)
plt.xlabel('Precision')
plt.ylabel('Recall')
plt.show()
def kfold():
model = joblib.load('model.pkl')
vectorizer = joblib.load('vectorizer.pkl')
data_x, data_y = make_data_x_y()
data_x = vectorizer.transform(data_x)
data_y = np.array(data_y)
scoring = {
'accuracy': 'accuracy',
"precision": "precision",
"recall": "recall",
"f1": "f1"
}
skf = StratifiedKFold(n_splits=5, shuffle=True)
scores = cross_validate(model, data_x, data_y, cv=skf, scoring=scoring)
for key, value in scores.items():
print(f'{key} : {value.mean()}')