def add_twit_user():
if request.method == "POST":
result = request.form
username = result["username"]
api = twitter_api()
users = api.get_user(screen_name=username)
# tweets = api.user_timeline(screen_name = username, count=300,
# include_rts = False, exclude_replies=True)
# tweets = api.user_timeline(screen_name = username, tweet_mode ="extend")
db_users = Users()
db_users.id = users.id
db_users.username = users.screen_name
db_users.full_name = users.name
db_users.followers = users.followers_count
db.session.add(db_users)
print('Users')
#tweet text
raw_tweets = api.user_timeline(users.screen_name,
count=300,
include_rts=False,
exclude_replies=True,
tweet_mode="extended")
print('raw_tweets')
# 해당 user가 트윗을 한 개 이상 한 경우에만 db에 저장
if len(raw_tweets) >= 1:
for tweet in raw_tweets:
en = EmbeddingClient(host='54.180.124.154', port=8989)
one_tweet = [tweet.full_text]
print('one_tweet')
embedding_result = en.encode(texts=one_tweet)
print('embedding_result')
insert_tweet = Tweet(id=tweet.id,
text=tweet.full_text,
embedding=embedding_result[0],
user_id=users.id)
db.session.add(insert_tweet)
db.session.commit()
return render_template('add_routesdd.html')
def analyze():
users = Users.query.all()
text = []
id = []
prediction = 0
if request.method == "POST":
print(dict(request.form))
result = request.form
# import all datas from the table
for user in users:
tweets = Tweet.query.with_entities(
Tweet.embedding).filter(Tweet.user_id == user.id).all()
for tweet in tweets:
append_to_with_label(text, tweet, id, user.id)
# # 3D array to 2D array
# text_array = np.array(text)
# nsamples, nx, ny = text_array.shape
# text_2d = text_array.reshape(nsamples, nx * ny)
# Model import
if os.path.isfile(FILEPATH):
en = EmbeddingClient(host='54.180.124.154', port=8989)
model = pickle.load(open('model.pkl', 'rb'))
pred_id = model.predict(en.encode(texts=[result['text']]))
prediction = int(pred_id[0])
else:
model = LogisticRegression(warm_start=True)
model.fit(text, id)
pred_id = model.predict(en.encode(texts=[result['text']]))
prediction = int(pred_id[0])
pickle.dump(model, open('model.pkl', 'wb'))
# Predction result
pred_res = Users.query.filter(Users.id == prediction).first()
return render_template('analytics.html', prediction=pred_res)
def analyze():
if request.method == 'POST':
users = Users.query.all()
# prediction = ""
# compare_text = ""
raw_user_1 = request.form["User1"]
raw_user_2 = request.form["User2"]
user_1 = Users.query.filter_by(id=raw_user_1).one()
user_2 = Users.query.filter_by(id=raw_user_2).one()
embedding = []
labels = []
for tw_1 in user_1.tweets:
embedding.append(tw_1.embedding)
labels.append(user_1.username)
for tw_2 in user_2.tweets:
embedding.append(tw_2.embedding)
labels.append(user_2.username)
classifier = RandomForestClassifier()
classifier.fit(embedding, labels)
compare_text = request.form['text']
en = EmbeddingClient(host='54.180.124.154', port=8989)
predict_embedding = en.encode(texts=[compare_text])
prediction = classifier.predict(predict_embedding)
print(f"Compare string {compare_text}")
print(f"Prediction Results {prediction}")
return render_template("analytics.html",
users=users,
predict=prediction,
compare_text=compare_text)
from flask_sqlalchemy import SQLAlchemy
from flask_migrate import Migrate
from embedding_as_service_client import EmbeddingClient
db = SQLAlchemy()
migrate = Migrate()
en = EmbeddingClient(host='54.180.124.154', port=8989)
# db 테이블 생성
class Users(db.Model):
id = db.Column(db.Integer, primary_key=True)
username = db.Column(db.Integer, nullable=False)
full_name = db.Column(db.String, nullable=False, unique=True)
followers = db.Column(db.Integer, nullable=False)
location = db.Column(db.String, nullable=False)
# tweet_id = db.Column(db.Integer, db.ForeignKey('Tweets.id'))
# tweet = db.relationship('Tweets', backref='users', lazy=True)
def __repr__(self):
return f"<User{self.id} {self.username}>"
class Tweets(db.Model):
id = db.Column(db.Integer, primary_key=True)
text = db.Column(db.String, nullable=False)
embedding = db.Column(db.PickleType)
user_id = db.Column(db.BigInteger, db.ForeignKey('users.id'))
users = db.relationship('Users', backref='tweets', lazy=True)
from embedding_as_service_client import EmbeddingClient
from typing import List
from sklearn.linear_model import LogisticRegression
en = EmbeddingClient(host="54.180.124.154", port=8989)
def get_embeddings(text_list: List[str]) -> List[List[float]]:
vecs = en.encode(texts=text_list)
return vecs
