class Model:
def __init__(self, name):
self.name = name
self.construct()
def construct(self):
if self.name == "linear":
self.model = LinearSVC()
elif self.name == "logistic":
# self.model = LogisticRegression(solver='lbfgs',multi_class='multinomial', n_jobs=-1)
self.model = LogisticRegression()
elif self.name == "cnn":
self.model = CNN()
elif self.name == "bayes":
self.model = MultinomialNB()
def predict(self, X):
return self.model.predict(X)
def predict_proba(self, X):
if self.name == "linear":
return self.model.decision_function(X)
elif self.name == "logistic":
return self.model.predict_proba(X)
elif self.name == "cnn":
return self.model.decision_function(X)
elif self.name == "bayes":
return self.model.predict_proba(X)
def fit(self, X, Y):
self.model.fit(X, Y)
def test(method, train_data, test_data):
train_text, train_label, train_votes = function.data2vec(train_data)
test_text, test_label, test_votes = function.data2vec(test_data)
# train
vec_model = CountVectorizer()
train_text_vec = vec_model.fit_transform(train_text)
if method == 'naive_bayes':
cls = MultinomialNB()
else:
cls = LinearSVC()
cls.fit(train_text_vec, train_label)
# predict label
test_text_vec = vec_model.transform(test_text)
predicted_label = cls.predict(test_text_vec)
print(
classification_report(test_label,
predicted_label,
digits=4,
target_names=EMOTIONS_LIST,
zero_division=0))
# predict prob for coefficient
if method == 'naive_bayes':
predicted_votes = cls.predict_proba(test_text_vec)
else:
predicted_votes = cls.decision_function(test_text_vec)
coefficient = 0
for p_vote, t_vote in zip(predicted_votes, test_votes):
coefficient += pearsonr(p_vote, t_vote)[0]
coefficient /= len(predicted_votes)
print(f"coefficient = {coefficient:.4f}")
df_cm = pd.DataFrame(conf_matrix, index=class_names, columns=class_names )
plt.figure(figsize=(5, 5))
hm = sns.heatmap(df_cm, cbar=False, annot=True, square=True, fmt='d', annot_kws={'size': 20},
yticklabels=df_cm.columns, xticklabels=df_cm.columns)
hm.yaxis.set_ticklabels(hm.yaxis.get_ticklabels(), rotation=0, ha='right', fontsize=10)
hm.xaxis.set_ticklabels(hm.xaxis.get_ticklabels(), rotation=0, ha='right', fontsize=10)
plt.ylabel('True label', fontsize=20)
plt.xlabel('Predicted label', fontsize=20)
# Show heat map
plt.tight_layout()
plt.show()
# -----------------------------------------------------------------------
# Plot ROC Area Under Curve
y_predict_probability = clf.decision_function(X_test)
fpr, tpr, _ = roc_curve(y_test, y_predict_probability)
auc = roc_auc_score(y_test, y_predict_probability)
# print(fpr)
# print(tpr)
# print(auc)
plt.figure()
lw = 2
plt.plot(fpr, tpr, color='darkorange',
lw=lw, label='ROC curve (area = %0.2f)' % auc)
plt.plot([0, 1], [0, 1], color='navy', lw=lw, linestyle='--')
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
