aateg

import nltk
nltk.download('machado')
from nltk.probability import FreqDist
from nltk.tokenize import word_tokenize
nltk.download('punkt')

# corpus dom casmurro
corpus_dom_casmurro = nltk.corpus.machado.raw('romance/marm08.txt')

# pre processamento

aateg

import nltk
nltk.download('stopwords')

# retorna lista de stopwords em portugues
stopwords = nltk.corpus.stopwords.words('portuguese')

aateg

import nltk

# id do corpus 
# no nosso caso estamos usando id machado
nltk_id = 'machado'

# eh necessario baixar o corpus
nltk.download(nltk_id)

# agora o corpus esta acessivel 

aateg

from sklearn.metrics import confusion_matrix

tn, fp, fn, tp = confusion_matrix(y_test, y_predito).ravel()

precision = tp/(tp + fp)
recall = tp/(tp + fn)
fpr = fp/(fp + tn)

aateg

from sklearn.metrics import roc_auc_score

"""
Função que calcula a área sob a curva ROC.
Recebe os valores preditos e os valores 
do teste para calcular a area. Seu retorno
é um valor no intervalo [0, 1].
"""

auc_score = roc_auc_score(y_true, y_scores)

aateg

from sklearn import metrics

"""
a função do sklearn retorna uma tupla
contendo numpy arrays com true positive rate (TPR)
false positive rate (FPR) e threshold
"""

fpr, tpr, thresholds = metrics.roc_curve(y_test, y_predito)

aateg

def nnTrain(epsilon, alpha, max_iter):
    input_layer_size = x_train.shape[1]
    hidden_layer_size = 800
    num_labels = 10
    theta_1, theta_2 = randomInit(input_layer_size, hidden_layer_size, num_labels)
    
    for i in range(max_iter):
      J_theta, Theta1_grad, Theta2_grad = nnRegCostFunction(
          theta_1, theta_2, x_train, y_train,

aateg

def classifications(theta_1, theta_2, X,):
  a1 = np.append(np.ones(shape=(X.shape[0], 1)), X, axis=1)

  z2 = a1 @ theta_1.transpose()
  a2 = sigmoid(z2)
  a2 = np.append(np.ones(shape=(a2.shape[0], 1)), a2, axis=1)

  z3 = a2 @ theta_2.transpose()
  a3 = sigmoid(z3)
  

aateg

def nnRegCostFunction(theta_1, theta_2, X, y, input_layer_size, hidden_layer_size, num_labels):    
  # Variáveis úteis (m=número de imagens, n=número de pixels por imagem)
  [m, n] = X.shape

  # Algoritmo Feedforward
  # Adiciona a bias unit (uma coluna de 1) ao dataset
  a1 = np.append(np.ones(shape=(X.shape[0], 1)), X, axis=1)

  z2 = a1 @ theta_1.transpose()
  a2 = sigmoid(z2)

aateg

def randomInit(input_layer_size, hidden_layer_size, num_labels):
  # Parâmetros da primeira camada (incluindo bias)
  # As matrizes têm formato (num_entradas, num_saidas)
  size1 = (hidden_layer_size, input_layer_size+1)
  theta1_ini = np.random.normal(0, .1, size=size1)

  # Parâmetros da primeira camada (incluindo bias)
  size2 = (num_labels, hidden_layer_size+1)
  theta2_ini = np.random.normal(0, .1, size=size2)