# -*- coding: utf-8 -*-
"""Copia de Modelo con MLP.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1Tz3vMBacQRtb-nGhUrwNjr9dI3LBy29_
"""

import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.svm import SVC
from sklearn import preprocessing
from sklearn.model_selection import train_test_split as split
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestClassifier
from sklearn.neural_network import MLPClassifier
from sklearn.model_selection import  GridSearchCV
from sklearn.metrics import classification_report as report
import numpy as np
import matplotlib.pyplot as ptl
from imblearn.over_sampling import SMOTE
from datasets import load_dataset

#from google.colab import drive
#drive.mount('/content/drive')

#from google.colab import drive
#drive.mount("/content/drive")

#file = "/content/drive/MyDrive/Corpus_cleaned.csv"
dataset = load_dataset("csv", data_files="Corpus_cleaned.csv")
df = pd.read_csv(dataset)
corpus = df.iloc[:, 1]
labels = df.iloc[:, 2]

print(corpus[0])
print(labels[0])

#Vectorización
transformer = TfidfVectorizer()
corpus = corpus.fillna(' ')
tfidf = transformer.fit_transform(corpus)
transformer.get_feature_names_out() # Palabras
X = tfidf.toarray()

le = preprocessing.LabelEncoder()
le.fit(labels)
y = le.transform(labels)
print(y[0])

#Para el balanceo de clases
smote = SMOTE(random_state=42)
X_sm, y_sm = smote.fit_resample(X, y)
print(f'''Cambio de X antes de SMOTE: {X.shape}
      Cambio de X despues de SMOTE: {X_sm.shape}''')

print('\nBalance positivo y negativo de la clases (%):')
unique, counts = np.unique(y_sm, return_counts=True)
print(counts*100)

#Usa el 80% para entrenamiento
X_train, X_test, y_train, y_test = split(X_sm, y_sm, train_size=0.8, random_state=42)

#Creación del modelo MLP
clf = MLPClassifier(max_iter=100, activation= 'tanh', alpha= 0.05, hidden_layer_sizes= (100,), learning_rate= 'adaptive', solver= 'adam')
clf.fit(X_train, y_train)

#Obtener predicciones
preds = clf.predict(X_test)
print(report(y_test, preds))

#test
tweet = "dia de la madre"
vectTweet = transformer.transform(np.array([tweet]))

prediction = clf.predict(vectTweet)
print('El tweet es', 'ciberbullying' if prediction[0]==0 else 'no' if prediction[0]==1 else 'otro')
print(prediction[0])