def post(self):
try:
formData = request.json
ingredients = [
x.strip() for x in formData['ingredients'].split(',')
]
X_count = vectorizer.transform(ingredients)
countvector = X_count.toarray().astype(int)
x_ingredients = np.array([sum(x) for x in zip(*countvector)])
cooking_steps = lemmatize_text(clean_text(
formData['cookingSteps']))
x_cooking = tfidf_vectorizer.transform([cooking_steps
]).toarray()[0]
x = np.concatenate([x_ingredients, x_cooking])
prediction = classifier.predict(x.reshape(1, -1))
types = {0: "Bad Recipe", 1: "Good Recipe"}
response = jsonify({
"statusCode":
200,
"status":
"Prediction made",
"result":
"Recipe Quality Prediction: " + types[prediction[0]]
})
response.headers.add('Access-Control-Allow-Origin', '*')
return response
except Exception as error:
return jsonify({
"statusCode": 500,
"status": "Could not make prediction",
"error": str(error)
})
def parse(self, content_list, image_list, feeder_running, parser_running):
name = current_process().name
self._logger.info(name, '크롬 드라이버 로딩 중..')
parser_driver = webdriver.Chrome(self._chromedriver)
parser_driver.implicitly_wait(5)
parser_driver.get('https://cyworld.com')
for cookie in self._cookie:
parser_driver.add_cookie(cookie)
self._logger.info(name, '크롬 드라이버 로딩 완료')
while feeder_running.value or len(content_list) != 0:
try:
if len(content_list) != 0:
# 공유 리스트에서 게시물 URL 추출 및 접속
target_url = content_list.pop(0)
self._logger.info(name, target_url)
parser_driver.get(target_url)
# 필요한 데이터 추출
date = parser_driver \
.find_element_by_css_selector('div.view1 p')
images = parser_driver \
.find_elements_by_css_selector('section.imageBox')
texts = parser_driver \
.find_elements_by_css_selector('section.textBox')
# 원본 제목
title = parser_driver \
.find_element_by_id('cyco-post-title') \
.get_attribute('innerText')
# 파일 저장을 위해 전처리한 제목 (파일명으로 사용됨)
preprocessed_title = to_valid_filename(title)
# 게시글 날짜 업로드 날짜
post_date = extract_date(date.get_attribute('innerText'))
# 게시글 데이터 병합
post_text = '[ {} ]\n\n'.format(title)
for text in texts:
current_text = text.get_attribute('innerText').strip()
if len(current_text):
post_text += clean_text(current_text) + '\n'
# 이미지 목록 추출
for image in images:
imgs = image.find_elements_by_tag_name('img')
for img in imgs:
src = update_size(img.get_attribute('src'))
image_list.append({
'title': preprocessed_title,
'date': post_date,
'content': post_text,
'src': src
})
self._logger.info(
name,
'{}_{} 포스트 파싱 됨'.format(post_date, title))
# 싸이월드 서버 부하 방지를 위해 잠시 대기
time.sleep(1)
except Exception as e:
self._logger.error(str(e))
parser_running.value = 0
parser_driver.close()
self._logger.info(name, '종료')
# count vectorizer
vectorizer = CountVectorizer()
X_count = vectorizer.fit_transform(df['recipe_ingredients'])
countvector = X_count.toarray().astype(int)
df['ingedients_vector'] = countvector.tolist()
# # tfidf
# tfidf_transformer = TfidfTransformer()
# X_tfidf = tfidf_transformer.fit_transform(X_count)
# df['ingedients_tfidf_vector'] = X_tfidf.toarray().tolist()
ingredients = vectorizer.get_feature_names()
# Cooking Steps
df['cooking_steps_all'] = df['cooking_steps'].apply(
lambda x: " ".join(ast.literal_eval(x)))
df['cooking_steps_clean'] = [clean_text(i) for i in df['cooking_steps_all']]
df['cooking_steps_lemma'] = [
lemmatize_text(i) for i in df['cooking_steps_clean']
]
tfidf_vectorizer, x_tfidf, features = tfidf(df['cooking_steps_lemma'])
df_cooking_steps = pd.DataFrame(x_tfidf.toarray(), columns=features)
df = pd.concat([df, df_cooking_steps], axis=1)
# Target
df['is_good_recipe'] = df['is_good_recipe'].astype(int)
# Split the dataset into features and labels
y = np.array(df['is_good_recipe'])
x_ingredients = np.array(
df['ingedients_vector'].tolist()) # ingedients_tfidf_vector
x_cooking = df[features].to_numpy()