import tweepy

# Configuració de les claus d'API
consumer_key = 'your_consumer_key'
consumer_secret = 'your_consumer_secret'
access_token = 'your_access_token'
access_token_secret = 'your_access_token_secret'

# Autenticació amb l'API de Twitter
auth = tweepy.OAuthHandler(consumer_key, consumer_secret)
auth.set_access_token(access_token, access_token_secret)
api = tweepy.API(auth)

# Funció per obtenir tweets
def obtenir_tweets(paraula_clau, nombre_tweets):
    tweets = tweepy.Cursor(api.search, q=paraula_clau, lang="ca").items(nombre_tweets)
    tweets_list = [[tweet.text, tweet.created_at, tweet.user.screen_name] for tweet in tweets]
    return tweets_list

# Exemple de recopilació de tweets
tweets = obtenir_tweets('Machine Learning', 100)

import re

def netejar_text(text):
    text = re.sub(r'http\S+', '', text)  # Eliminar enllaços
    text = re.sub(r'@\w+', '', text)  # Eliminar mencions
    text = re.sub(r'#\w+', '', text)  # Eliminar hashtags
    text = re.sub(r'\W', ' ', text)  # Eliminar caràcters especials
    text = text.lower()  # Convertir a minúscules
    return text

# Aplicar la funció de neteja
tweets_netejats = [netejar_text(tweet[0]) for tweet in tweets]

import nltk
from nltk.tokenize import word_tokenize
from nltk.corpus import stopwords
from nltk.stem import WordNetLemmatizer

nltk.download('punkt')
nltk.download('stopwords')
nltk.download('wordnet')

lemmatizer = WordNetLemmatizer()
stop_words = set(stopwords.words('catalan'))

def preprocessar_text(text):
    tokens = word_tokenize(text)
    tokens = [lemmatizer.lemmatize(word) for word in tokens if word not in stop_words]
    return ' '.join(tokens)

# Aplicar la funció de preprocessament
tweets_preprocessats = [preprocessar_text(tweet) for tweet in tweets_netejats]

from sklearn.feature_extraction.text import TfidfVectorizer

vectorizer = TfidfVectorizer(max_features=5000)
X = vectorizer.fit_transform(tweets_preprocessats).toarray()

# Exemple d'etiquetatge manual
y = [1 if 'positiu' in tweet else 0 for tweet in tweets_preprocessats]  # 1 per positiu, 0 per negatiu/neutre

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression

# Divisió de les dades en conjunts d'entrenament i prova
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0)

# Entrenament del model
model = LogisticRegression()
model.fit(X_train, y_train)

from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score

# Prediccions
y_pred = model.predict(X_test)

# Càlcul de mètriques
accuracy = accuracy_score(y_test, y_pred)
precision = precision_score(y_test, y_pred)
recall = recall_score(y_test, y_pred)
f1 = f1_score(y_test, y_pred)

print(f'Exactitud: {accuracy}')
print(f'Precisió: {precision}')
print(f'Recall: {recall}')
print(f'F1-score: {f1}')

from flask import Flask, request, jsonify

app = Flask(__name__)

@app.route('/predict', methods=['POST'])
def predict():
    data = request.get_json(force=True)
    text = data['text']
    text_clean = preprocessar_text(netejar_text(text))
    text_vector = vectorizer.transform([text_clean]).toarray()
    prediction = model.predict(text_vector)
    return jsonify({'sentiment': 'positiu' if prediction[0] == 1 else 'negatiu'})

if __name__ == '__main__':
    app.run(debug=True)