DanielDaCosta

import re
import nltk
from nltk.corpus import stopwords
from nltk.stem.wordnet import WordNetLemmatizer
def clean_text(text):
    text = text.lower()
    
    # '@' mention. Even tough @ adds some information to the message, 
    # this information doesn't add value build the classifcation model
    text = re.sub(r'@[A-Za-z0-9_]+','', text)

DanielDaCosta

from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
vocabulary_size = 20000 # Choosing size of vocabulary
tokenizer = Tokenizer(num_words=vocabulary_size)
tokenizer.fit_on_texts(df['message'])
sequences = tokenizer.texts_to_sequences(df['message'])
# Pads sequences to the same length: MAXLEN
MAXLEN = 50
X = pad_sequences(sequences, maxlen=MAXLEN)
y = df[output_columns_all]

DanielDaCosta

main_input = Input(shape=(MAXLEN,), dtype='int32', name='main_input')
x = Embedding(input_dim=vocabulary_size, output_dim=50, input_length=MAXLEN)(main_input)
x = Dropout(0.3)(x)
x = Conv1D(64, 5, activation='relu')(x)
x = MaxPooling1D(pool_size=4)(x)
x = LSTM(100)(x)
x = Dropout(0.3)(x)

DanielDaCosta

output_array = [] 
metrics_array = {}
loss_array = {}
for i, dense_layer in enumerate(output_columns_binary):
    name = f'binary_output_{i}' 
    # A Dense Layer is created for each output
    binary_output = Dense(1, activation='sigmoid', name=name)(x)
    output_array.append(binary_output)
    metrics_array[name] = 'binary_accuracy'
    loss_array[name] = 'binary_crossentropy'

DanielDaCosta

categorical_output = Dense(3, activation='softmax', name='categorical_output')(x)
output_array.append(categorical_output)
metrics_array['categorical_output'] = 'sparse_categorical_accuracy'
loss_array['categorical_output'] = 'sparse_categorical_crossentropy'

DanielDaCosta

weight_binary = {0: 0.5, 1: 7} #values obtained through calculations
weight_categorical = {0: 1.4, 1: 0.43, 2: 7}
classes_weights = {}
# There are 35 outputs in output_columns_binary, having classes (0 or 1)
for i, dense_layer in enumerate(output_columns_binary):
    name = f'binary_output_{i}'
    classes_weights[name] = weight_binary
# There is only 1 output in output_columns_categorical, having classes (0, 1 or 2)
for i, dense_layer in enumerate(output_columns_categorical):
    name = 'categorical_output'

DanielDaCosta

model = Model(inputs=main_input, outputs=output_array)
model.compile(optimizer='adadelta',
              loss=loss_array,
              metrics=metrics_array)
model.fit(X_train, y_train_output,
          epochs=40, batch_size=512,
         class_weight=classes_weights, verbose=0);

DanielDaCosta

variable "environment" {
    description = "Env"
    default     = "dev"
}

variable "name" {
    description = "Application Name"
    type        = string
}

DanielDaCosta

resource "aws_sqs_queue" "queue" {
  name                      = "apigateway-queue"
  delay_seconds             = 0
  max_message_size          = 262144
  message_retention_seconds = 86400
  receive_wait_time_seconds = 10

  tags = {
    Product = local.app_name
  }

DanielDaCosta

{
    "Version": "2012-10-17",
    "Statement": [
      {
        "Effect": "Allow",
        "Action": [
          "logs:CreateLogGroup",
          "logs:CreateLogStream",
          "logs:DescribeLogGroups",
          "logs:DescribeLogStreams",