Apertando_o_parafuso_certo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Naive Bayes "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Análise de Sentimentos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import nltk\n",
    "import re\n",
    "import pandas as pd\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "from sklearn import metrics\n",
    "from sklearn.model_selection import cross_val_predict"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![title](tweets.jpeg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Ler arquivo de dados e conta a quantidade de linhas**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = pd.read_csv('Tweets_Mg.csv',encoding='utf-8')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 8199 entries, 0 to 8198\n",
      "Data columns (total 26 columns):\n",
      "Unnamed: 0                   8199 non-null int64\n",
      "Created At                   8199 non-null object\n",
      "Text                         8199 non-null object\n",
      "Geo Coordinates.latitude     104 non-null float64\n",
      "Geo Coordinates.longitude    104 non-null float64\n",
      "User Location                5489 non-null object\n",
      "Username                     8199 non-null object\n",
      "User Screen Name             8199 non-null object\n",
      "Retweet Count                8199 non-null int64\n",
      "Classificacao                8199 non-null object\n",
      "Observação                   1 non-null object\n",
      "Unnamed: 10                  0 non-null float64\n",
      "Unnamed: 11                  0 non-null float64\n",
      "Unnamed: 12                  0 non-null float64\n",
      "Unnamed: 13                  0 non-null float64\n",
      "Unnamed: 14                  0 non-null float64\n",
      "Unnamed: 15                  0 non-null float64\n",
      "Unnamed: 16                  0 non-null float64\n",
      "Unnamed: 17                  0 non-null float64\n",
      "Unnamed: 18                  0 non-null float64\n",
      "Unnamed: 19                  0 non-null float64\n",
      "Unnamed: 20                  0 non-null float64\n",
      "Unnamed: 21                  0 non-null float64\n",
      "Unnamed: 22                  0 non-null float64\n",
      "Unnamed: 23                  0 non-null float64\n",
      "Unnamed: 24                  0 non-null float64\n",
      "dtypes: float64(17), int64(2), object(7)\n",
      "memory usage: 1.6+ MB\n"
     ]
    }
   ],
   "source": [
    "dataset.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0     ���⛪ @ Catedral de Santo Antônio - Governador Valadares/MG https://t.co/JSbKamIqUJ                                                          \n",
       "1     � @ Governador Valadares, Minas Gerais https://t.co/B3ThIDJCSf                                                                              \n",
       "2     �� @ Governador Valadares, Minas Gerais https://t.co/dPkgzVR2Qw                                                                             \n",
       "3     ��� https://t.co/BnDsO34qK0                                                                                                                 \n",
       "4     ��� PSOL vai questionar aumento de vereadores e prefeito de BH na Justiça - Politica - Estado de Minas https://t.co/DMg7BGsek5              \n",
       "5     \" bom é bandido morto\"\\nDeputado Cabo Júlio é condenado e fica inelegível por 10 anos - Politica - Estado de Minas https://t.co/3GfAqvrFHS  \n",
       "6     \"..E 25% dos mineiros dizem não torcer para time nenhum,mesmo dentro de um estado com Atlético-MG e Cruzeiro. Pq?..\" https://t.co/fN5evlLQsR\n",
       "7     \"A gigantesca barba do mal\" em destaque no caderno Cultura do Estado de Minas. https://t.co/PxNurIkTlw                                      \n",
       "8     \"BB e governo de Minas travam disputa sobre depósitos judiciais\" https://t.co/CnMu2A2Qo5                                                    \n",
       "9     \"com vcs bh fica pequena!\" Belo Horizonte (pron. [bɛloɾiˈzõntʃi][10]) é a capital do estado de MG, com uma área de aproximadamente 331 km²  \n",
       "10    \"Daí a gente visita governador valadares\"\\n\"Que lugar eh esse mãe?\"\\n\"MG, a gente tem uns parente la\"\\n\\nMANO onde q eu NÃO tenho parente   \n",
       "11    \"É bonita e é bonita...\" \\n#latergram #ibituruna #home @ Governador Valadares, Minas Gerais https://t.co/y12TtNqpnH                         \n",
       "12    \"erro desconhecido\" é mato! Aliás, é da secretaria estadual de fazenda que tá assim... MG não precisa de dinheiro não, Governador?          \n",
       "13    \"La La Land: Cantando Estações\" arrasa no Globo de Ouro - Estado de Minas https://t.co/aQ0NG1bp5f https://t.co/aY3qlrVnTE                   \n",
       "14    \"La La Land: Cantando Estações\" arrasa no Globo de Ouro - Estado de Minas https://t.co/qRI9DCiicc                                           \n",
       "15    \"Los abusos, aun en el estado más sólido, son minas sordas que tarde o temprano estallan\" Joseph Sanial-Dubay                               \n",
       "16    \"Mesmo sem muito dinheiro no caixa o governo de Minas Gerais compra 2 (DOIS) helicópteros.\" Dois... Helicópteros... Minas Gerais... #2017 s1\n",
       "17    \"Modelo de segurança será o nosso legado\", diz Cláudio Beato - Politica - Estado de Minas https://t.co/1MqOJo7Nnf                           \n",
       "18    \"Modelo de segurança será o nosso legado\", diz Cláudio Beato - Politica - Estado de Minas https://t.co/kA09mEn4xN                           \n",
       "19    \"Modelo de segurança será o nosso legado\", diz Cláudio Beato - Politica - Estado de Minas https://t.co/S97Khxldiu                           \n",
       "20    \"O governo de Minas Gerais não comentou o assunto. \"\\n\\nhttps://t.co/SNYMRfQ737                                                             \n",
       "21    \"O governo federal publicou no dia 27 de dezembro de 2016, a Medida Provisória 764/16 que permite a diferenciação... https://t.co/0DlDpZ9WGz\n",
       "22    \"Polícia Federal acaba de divulgar que encontrou provas no apartamento do Lula, mas só até a 4ª série.\" Jornal Estado de Minas              \n",
       "23    \"Praia da Estação\" lota neste sábado de verão em BH - Gerais - Estado de Minas https://t.co/DpG3CNfZBb                                      \n",
       "24    \"Representando mais uma ferramenta de monitoramento da educação da rede estadual em Minas, a Secretaria de Estado... https://t.co/brsfBqHXlQ\n",
       "25    \"uso de aeronave pertencente ao  Estado de MG, em domingo, p/ fins pessoais,  em claro desvio de finalidade\" - https://t.co/WEfv5iLyju      \n",
       "26    \"vou postar umas curiosidades sobre o corpo humano\"\\n\\n*o Japão é menor que o estado de minas gerais* ????¿¿¿.?¿                            \n",
       "27    #ACORDAMINAS!!!\\n\\nO governador Fernando Pimentel, depois de ser flagrado usando um helicóptero do Estado de MG... https://t.co/bedW4w4jij  \n",
       "28    #ACORDAMINAS!!!!\\n\\nO povo mineiro tem que da um pé na bunda  desse  canalha... https://t.co/ZKZNVmyXbg                                     \n",
       "29    #Alerta - Nome da CEEE é utilizado em golpe no Estado. https://t.co/i5ndsWhXR4 https://t.co/yaEM6s3Z2f                                      \n",
       "                                                       ...                                                                                        \n",
       "70    #i5gornascimento @i5gornascimento @cristiano UFJF abre 62 vagas para cursos em Governador Valadares https://t.co/NnA9YFp4BP                 \n",
       "71    #i5gornascimento @i5gornascimento g1 Corpo em estado de decomposição é encontrado em Poços de Caldas, MG https://t.co/ndVVD6kdly            \n",
       "72    #i5gornascimento @i5gornascimento g1 Prefeito de Aimorés decreta estado de emergência administrativa https://t.co/0Jh7ARbSnk                \n",
       "73    #i5gornascimento @i5gornascimento g1 Vereadores de Governador Valadares não terão 13º salário em 2017 https://t.co/9dJnmDxEAI               \n",
       "74    #i5gornascimento @i5gornascimento Governador Valadares registra dois assassinatos em menos de uma hora https://t.co/ZcuZugp8dq              \n",
       "75    #i5gornascimento @i5gornascimento UFJF abre 62 vagas para cursos em Governador Valadares https://t.co/FYeEFvieCy                            \n",
       "76    #ImpeachmentPimentel #PTNuncaMais \\nMesmo em calamidade financeira, governo de MG compra mais dois helicópteros https://t.co/rcLSmbvdoH     \n",
       "77    #ImpeachmentPimentel #PTNuncaMais #ForaPT #ForaPCdoB #ForaPSOL #ForaRede \\nhttps://t.co/0RRGhWdmdB                                          \n",
       "78    #Janeiro, #PartiuFerias, GOVERNADOR VALADARES/MG, Pousada Vale Silvestre EcoPark, Comprar: https://t.co/rNRqRFbUeM… https://t.co/2UyZ6Ojdp7 \n",
       "79    #Janeiro, #PartiuFerias, GOVERNADOR VALADARES/MG, Pousada Vale Silvestre EcoPark, Comprar: https://t.co/rNRqRFbUeM… https://t.co/qw78x6qm7k \n",
       "80    #jornaldacultura\\nPor Favor comentem a compra de dois helicópteros pelo governo de Minas Gerais e o uso do Helicópteros por Pimentel        \n",
       "81    #METADEdoBRASILpertenceaosPOLÍTICOSaoutraédosBANCOSeEMPRESÁRIOS  Politica - Estado de Minas https://t.co/usvOMucmuq                         \n",
       "82    #Microsoft  #lumia galera hj descobrir que aqui n estado do rio de janeiro n tem autorizada .(S.PAULO E MINAS ) NAO COMPREI WINDOWS PHONE   \n",
       "83    #Mídia:  Press Release from Business Wire : Consumer Technology ... - Estado de Minas https://t.co/FGoxtwksdM                               \n",
       "84    #Mídia:  Press Release from Business Wire : Digital Living Network Alliance - Estado de Minas https://t.co/wj7wMchv8S                       \n",
       "85    #Mídia:  Press Release from Business Wire : IBC - Estado de Minas https://t.co/Uy5QuJo8wx                                                   \n",
       "86    #MINAS Postos dos Sine oferecem mais de 1.500 vagas de emprego no estado https://t.co/ZTrZ8MGrRp #Trabalho #Emprego #MG #Oportunidades      \n",
       "87    #Minas vai dobrar oferta de trabalho e estudo nos #presídios do Estado\\nhttps://t.co/CC6gpVFTCX                                             \n",
       "88    #Minas vai dobrar oferta de trabalho e estudo nos #presídios do Estado\\nhttps://t.co/UxfPtAKxte                                             \n",
       "89    #Mortalidade por câncer cai 25% em 25 anos nos Estados Unidos - Estado de Minas https://t.co/fsZg1D8WCL                                     \n",
       "90    #musicagratis Fundação Hospitalar do Estado de Minas Gerais (FHEMIG): A Fundação Hospitalar…… https://t.co/JXxZCenikX                       \n",
       "91    #musicagratis Ministério Público de Minas Gerais (Araxá): O Ministério Público do Estado de…… https://t.co/4WqHRHwxOg                       \n",
       "92    #musicagratis Prefeitura de Rodeiro: A Prefeitura de Rodeiro, no Estado de Minas Gerais,… https://t.co/kdGlkcQKcG… https://t.co/NyF4ZQiZqm  \n",
       "93    #noticiasdorenan Em calamidade financeira, governo de MG compra dois helicópteros: O… https://t.co/oW6G9hElta                               \n",
       "94    #noticiasdorenan Governo de Minas Gerais reconhece Folias de Reis como Patrimônio Imaterial… https://t.co/AL18Rr45X3                        \n",
       "95    #noticiasdorenan Uso de depósitos judiciais pelo governo tem nova polêmica em MG: O Banco… https://t.co/kHuZpKQaj5                          \n",
       "96    #OperacaoBetaLab #TimBeta #Beta #BetaLab #BetaAjudaBeta #BetaSegueBeta UOL: Desastre ambiental em MG: Empresas ped… https://t.co/kwyeKjhtcI \n",
       "97    #OperacaoBetaLab #TimBeta #Beta #BetaLab #BetaAjudaBeta #BetaSegueBeta UOL: Governo em calamidade financeira: BB a… https://t.co/1rWaHUACqt \n",
       "98    #OperacaoBetaLab #TimBeta #Beta #BetaLab #BetaAjudaBeta #BetaSegueBeta UOL: Polêmica com filho de Pimentel: Govern… https://t.co/E1EhSA9Ti3 \n",
       "99    #operacaobetalab BB avisa Justiça que governo de MG não tem recursos em conta judicial https://t.co/W6AoFMMsvA #rt #sdv                     \n",
       "Name: Text, Length: 100, dtype: object"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Expante o tamanho de exibição de coluna de texto\n",
    "pd.set_option('display.max_colwidth', -1)\n",
    "dataset.Text.head(100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Excluir Linhas duplicadas**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset.drop_duplicates(['Text'],inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Unnamed: 0                   5765\n",
       "Created At                   5765\n",
       "Text                         5765\n",
       "Geo Coordinates.latitude     104 \n",
       "Geo Coordinates.longitude    104 \n",
       "User Location                3851\n",
       "Username                     5765\n",
       "User Screen Name             5765\n",
       "Retweet Count                5765\n",
       "Classificacao                5765\n",
       "Observação                   1   \n",
       "Unnamed: 10                  0   \n",
       "Unnamed: 11                  0   \n",
       "Unnamed: 12                  0   \n",
       "Unnamed: 13                  0   \n",
       "Unnamed: 14                  0   \n",
       "Unnamed: 15                  0   \n",
       "Unnamed: 16                  0   \n",
       "Unnamed: 17                  0   \n",
       "Unnamed: 18                  0   \n",
       "Unnamed: 19                  0   \n",
       "Unnamed: 20                  0   \n",
       "Unnamed: 21                  0   \n",
       "Unnamed: 22                  0   \n",
       "Unnamed: 23                  0   \n",
       "Unnamed: 24                  0   \n",
       "dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset.count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Removendo Colunas desnecessárias **"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Colunas Deletadas: Unnamed: 0\n",
      "Colunas Deletadas: Unnamed: 10\n",
      "Colunas Deletadas: Unnamed: 11\n",
      "Colunas Deletadas: Unnamed: 12\n",
      "Colunas Deletadas: Unnamed: 13\n",
      "Colunas Deletadas: Unnamed: 14\n",
      "Colunas Deletadas: Unnamed: 15\n",
      "Colunas Deletadas: Unnamed: 16\n",
      "Colunas Deletadas: Unnamed: 17\n",
      "Colunas Deletadas: Unnamed: 18\n",
      "Colunas Deletadas: Unnamed: 19\n",
      "Colunas Deletadas: Unnamed: 20\n",
      "Colunas Deletadas: Unnamed: 21\n",
      "Colunas Deletadas: Unnamed: 22\n",
      "Colunas Deletadas: Unnamed: 23\n",
      "Colunas Deletadas: Unnamed: 24\n"
     ]
    }
   ],
   "source": [
    "for i in dataset.columns.values:\n",
    "    if i.startswith('Unnamed'):\n",
    "        dataset.drop(i, axis=1, inplace=True)\n",
    "        print ('Colunas Deletadas:', i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Created At', 'Text', 'Geo Coordinates.latitude',\n",
       "       'Geo Coordinates.longitude', 'User Location', 'Username',\n",
       "       'User Screen Name', 'Retweet Count', 'Classificacao', 'Observação'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Visualizando o dataset\n",
    "dataset.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Distribuição das classes neutros, positivos e negativos**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f40953e4d68>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "dataset.Classificacao.value_counts().plot(kind='bar')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "** Separando tweets e suas classes\n",
    "**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "tweets = dataset['Text']\n",
    "classes = dataset['Classificacao']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "**Gerando o modelo**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "vectorizer = CountVectorizer(analyzer=\"word\")\n",
    "freq_tweets = vectorizer.fit_transform(tweets)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Cria um modelo usando Naive Bayes Multinomial**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=1.0, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "modelo = MultinomialNB()\n",
    "modelo.fit(freq_tweets,classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Avaliação o modelo ##"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Usaremos a técnica de Cross Validation para avaliação dos modelos.\n",
    "* Esta técnica será usada com 10 folds."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "**Fazendo o cross validation do modelo**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "resultados = cross_val_predict(modelo, freq_tweets, classes, cv=10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "** Métricas de validação do modelo**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "   Positivo       0.96      0.88      0.91      2840\n",
      "   Negativo       0.77      0.89      0.82       951\n",
      "     Neutro       0.81      0.84      0.82      1974\n",
      "\n",
      "avg / total       0.87      0.87      0.87      5765\n",
      "\n"
     ]
    }
   ],
   "source": [
    "sentimento=['Positivo','Negativo','Neutro']\n",
    "print (metrics.classification_report(classes,resultados,sentimento))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Apertando o parafuso certo**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "MultinomialNB(alpha=5, class_prior=None, fit_prior=True)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vectorizer = CountVectorizer(analyzer=\"word\")\n",
    "freq_tweets = vectorizer.fit_transform(tweets)\n",
    "modelo = MultinomialNB(alpha=5)\n",
    "modelo.fit(freq_tweets,classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "freq_testes = vectorizer.transform(testes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['Neutro', 'Neutro', 'Negativo', 'Neutro', 'Neutro'], dtype='<U8')"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fazendo a classificação com o modelo treinado.\n",
    "modelo.predict(freq_testes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "resultados = cross_val_predict(modelo, freq_tweets, classes, cv=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "   Positivo       0.89      0.92      0.90      2840\n",
      "   Negativo       0.84      0.78      0.81       951\n",
      "     Neutro       0.79      0.79      0.79      1974\n",
      "\n",
      "avg / total       0.85      0.85      0.85      5765\n",
      "\n"
     ]
    }
   ],
   "source": [
    "sentimento=['Positivo','Negativo','Neutro']\n",
    "print (metrics.classification_report(classes,resultados,sentimento))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}