padronizando criacao de pkl

api/NbEmotionsModel.py

@@ -1,18 +1,17 @@
 import pandas as pd
 import pickle
 
-# bag of words
-from sklearn.feature_extraction.text import TfidfVectorizer
+def news_prediction(texts):
+ model_file = "./models/emotion_pipeline.pkl"
+ try:
+ # Carregando o pipeline do arquivo .pkl
+ with open(model_file, 'rb') as model_file:
+ pipeline = pickle.load(model_file)
 
-#tfidf_vectorizer = TfidfVectorizer(use_idf=True)
+ # Fazendo previsões para os textos
+ predictions = pipeline.predict([texts])
 
-def make_prediction(my_sentence):
- with open("./models/nb_emotion.pkl", "rb") as f:
- model = pickle.load(f)
+ return predictions[0]
 
- with open("./models/tfidf_vectorizer_em.pkl", 'rb') as f:
- tfidf_vectorizer = pickle.load(f)
-
- new_sentence = tfidf_vectorizer.transform([my_sentence])
- prediction = model.predict(new_sentence)
- return prediction[0]
+ except Exception as e:
+ return str(e)

api/models/nb_emotions.py

@@ -1,5 +1,9 @@
 import pandas as pd
 import pickle
+from sklearn.pipeline import make_pipeline
+from sklearn.feature_extraction.text import CountVectorizer
+from sklearn.feature_extraction.text import TfidfVectorizer
+
 
 # bag of words
 from sklearn.feature_extraction.text import TfidfVectorizer
@@ -10,15 +14,12 @@
 df = pd.read_csv('../training_df/tweet_emotions.csv')
 train_data, test_data, train_target, test_target = train_test_split(df["content"], df["sentiment"], test_size=0.2, shuffle=True)
 
-tfidf_vectorizer = TfidfVectorizer(use_idf=True)
-X_train_vectors_tfidf = tfidf_vectorizer.fit_transform(train_data)
-X_test_vectors_tfidf = tfidf_vectorizer.transform(test_data)
-
-nb_tfidf = MultinomialNB(alpha = 0)
-nb_tfidf.fit(X_train_vectors_tfidf, train_target)
+# Criando um pipeline com o vetorizador TF-IDF e o classificador Multinomial Naive Bayes
+pipeline = make_pipeline(TfidfVectorizer(), MultinomialNB())
 
-with open("nb_emotion.pkl", "wb") as f:
- pickle.dump(nb_tfidf, f)
+# Ajustando o modelo ao conjunto de treinamento
+pipeline.fit(train_data, train_target)
 
-with open("tfidf_vectorizer_em.pkl", "wb") as f:
- pickle.dump(tfidf_vectorizer, f)
+# Salvando o pipeline em um arquivo .pkl
+with open("emotion_pipeline.pkl", "wb") as model_file:
+ pickle.dump(pipeline, model_file)