Modelos 100dim.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import sys\n",
    "import warnings\n",
    "#from torchtext import data\n",
    "\n",
    "if not sys.warnoptions:\n",
    "    warnings.simplefilter(\"ignore\")\n",
    "    \n",
    "import numpy as np\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import math, copy, time\n",
    "from torch.autograd import Variable\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "seaborn.set_context(context=\"talk\")\n",
    "%matplotlib inline\n",
    "\n",
    "from keras import backend as K\n",
    "import keras\n",
    "import os, re, sys\n",
    "import pandas as pd\n",
    "from keras.utils import to_categorical\n",
    "import ast\n",
    "from sklearn.utils import class_weight\n",
    "from keras.preprocessing import sequence\n",
    "from keras.models import load_model\n",
    "import torch.nn.functional as F\n",
    "from sklearn.metrics import accuracy_score, f1_score, precision_score, recall_score\n",
    "\n",
    "from nltk.corpus import stopwords\n",
    "from bs4 import BeautifulSoup\n",
    "from nltk import WordNetLemmatizer\n",
    "import itertools\n",
    "from collections import Counter\n",
    "from os import listdir\n",
    "from os.path import isfile, join\n",
    "import  json\n",
    "from sklearn.utils import class_weight\n",
    "\n",
    "\n",
    "from  keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation, Flatten, BatchNormalization\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras.layers import Embedding\n",
    "from keras.layers import Conv1D, MaxPooling1D, Input\n",
    "from keras.preprocessing import sequence\n",
    "from keras.models import Model\n",
    "from keras.layers import LSTM,GRU,Dense\n",
    "from keras.utils import Sequence,to_categorical\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Generación de data Train/Val/test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "stoplist = stopwords.words('english')\n",
    "lemmatizer=WordNetLemmatizer()\n",
    "\n",
    "emoji_pattern = re.compile(\"[\"\n",
    "        u\"\\U0001F600-\\U0001F64F\"  # emoticons\n",
    "        u\"\\U0001F300-\\U0001F5FF\"  # symbols & pictographs\n",
    "        u\"\\U0001F680-\\U0001F6FF\"  # transport & map symbols\n",
    "        u\"\\U0001F1E0-\\U0001F1FF\"  # flags (iOS)\n",
    "                           \"]+\", flags=re.UNICODE)\n",
    "\n",
    "def cleaner(post):       \n",
    "    s= re.sub(r\"http\\S+\", \" <url> \", post)   ##########elimina http    \n",
    "    s= re.sub(r'#\\w+ ?', \" <hashtag> \", s) ########3\n",
    "    s= re.sub(r'@\\w+ ?', \" <user> \", s)##############    \n",
    "    s= s.lower()\n",
    "    s = re.sub(r\"-\", \" - \", s)\n",
    "    s = re.sub(r\"\\(\", \" \", s)\n",
    "    s = re.sub(r\"\\)\", \" \", s)\n",
    "    s = re.sub(r\"\\?\", \" ? \", s)    \n",
    "    s = re.sub(r\"\\/\", \" \", s)\n",
    "    s = re.sub(r\"' \", \" ' \", s)\n",
    "    s = re.sub(r\" '\", \" ' \", s)\n",
    "    s = re.sub(r\"\\!\", \" ! \", s)\n",
    "    s=re.sub(\"[\\.]+\", \" . \", s)\n",
    "    s=re.sub(\"[\\,]+\", \" , \", s)\n",
    "    s=re.sub(\"[\\;]+\", \" ; \", s)\n",
    "    s=re.sub(\"[\\:]+\", \" : \", s)\n",
    "    s=re.sub('[\\\"]+', ' \" ', s)\n",
    "    s=re.sub(r'\\b[0-9]\\b', \" <number> \",  s)\n",
    "    s=re.sub(r'\\b[0-9][0-9]\\b', \" <number> \",  s)\n",
    "    s=re.sub(r'\\b[0-9][0-9][0-9]\\b', \" <number> \",  s)\n",
    "    s=re.sub(r'\\b[0-9][0-9][0-9][0-9]\\b', \" <number> \",  s)\n",
    "    s=emoji_pattern.sub(r'', s)\n",
    "    s= list(s.split())\n",
    "    s=[lemmatizer.lemmatize(token) for token in s]\n",
    "        \n",
    "    return s"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>tweet_content</th>\n",
       "      <th>harassment</th>\n",
       "      <th>IndirectH</th>\n",
       "      <th>PhysicalH</th>\n",
       "      <th>SexualH</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>pics mexican school girl naked melanie safka f...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>RT shereiqns Having curly hair is a gotdamn ch...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>BreakingNews 7 Unexplained Prisoners Deaths M...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>girl dirty quarterback latino pokemon young h...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>miliondollameat sexting n ngirl daddy n nme l...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0                                      tweet_content  harassment  \\\n",
       "0           0  pics mexican school girl naked melanie safka f...           1   \n",
       "1           1  RT shereiqns Having curly hair is a gotdamn ch...           0   \n",
       "2           2   BreakingNews 7 Unexplained Prisoners Deaths M...           0   \n",
       "3           3   girl dirty quarterback latino pokemon young h...           1   \n",
       "4           4   miliondollameat sexting n ngirl daddy n nme l...           1   \n",
       "\n",
       "   IndirectH  PhysicalH  SexualH  \n",
       "0          0          0        1  \n",
       "1          0          0        0  \n",
       "2          0          0        0  \n",
       "3          0          0        1  \n",
       "4          0          0        1  "
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train=pd.read_csv(\"../Train_data_compeition.csv\")\n",
    "df_val=pd.read_csv(\"../Validation_data_competition.csv\")\n",
    "df_test=pd.read_csv(\"../testset-competition.csv\")\n",
    "df_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2123, 5)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>tweet_content</th>\n",
       "      <th>harassment</th>\n",
       "      <th>IndirectH</th>\n",
       "      <th>PhysicalH</th>\n",
       "      <th>SexualH</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6374</td>\n",
       "      <td>RT GLOSSPHOENIX some of y all never watched co...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6375</td>\n",
       "      <td>OMG CAN T BELIEVE I M DOING THIS Selling my Dr...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6376</td>\n",
       "      <td>asa akira ava lauren beach sex free porn video...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6377</td>\n",
       "      <td>Anyone sassy smart mouth cuuuuteeee old favor...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6378</td>\n",
       "      <td>This is why you re scared of Karma you got to ...</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0                                      tweet_content  harassment  \\\n",
       "0        6374  RT GLOSSPHOENIX some of y all never watched co...           0   \n",
       "1        6375  OMG CAN T BELIEVE I M DOING THIS Selling my Dr...           0   \n",
       "2        6376  asa akira ava lauren beach sex free porn video...           1   \n",
       "3        6377   Anyone sassy smart mouth cuuuuteeee old favor...           0   \n",
       "4        6378  This is why you re scared of Karma you got to ...           1   \n",
       "\n",
       "   IndirectH  PhysicalH  SexualH  \n",
       "0          0          0        0  \n",
       "1          0          0        0  \n",
       "2          0          0        1  \n",
       "3          0          0        0  \n",
       "4          0          0        1  "
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_val.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>tweet_content</th>\n",
       "      <th>Harassment</th>\n",
       "      <th>IndirectH</th>\n",
       "      <th>PhysicalH</th>\n",
       "      <th>SexualH</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>is it wrong what about mezcal</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>is mkr on texts the sound you make with incaka...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>is on now go girl bre</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>is only beaten by on anzac day eve k share which</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>is that a bowl cut on that chick</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                       tweet_content  Harassment  IndirectH  \\\n",
       "0                      is it wrong what about mezcal         NaN        NaN   \n",
       "1  is mkr on texts the sound you make with incaka...         NaN        NaN   \n",
       "2                              is on now go girl bre         NaN        NaN   \n",
       "3   is only beaten by on anzac day eve k share which         NaN        NaN   \n",
       "4                   is that a bowl cut on that chick         NaN        NaN   \n",
       "\n",
       "   PhysicalH  SexualH  \n",
       "0        NaN      NaN  \n",
       "1        NaN      NaN  \n",
       "2        NaN      NaN  \n",
       "3        NaN      NaN  \n",
       "4        NaN      NaN  "
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "conteo de tweets clase Harassment Counter({0: 3661, 1: 2713})\n",
      "conteo de tweets clase IndirectH Counter({0: 6319, 1: 55})\n",
      "conteo de tweets clase PhysicalH Counter({0: 6298, 1: 76})\n",
      "conteo de tweets clase SexualH Counter({0: 3792, 1: 2582})\n"
     ]
    }
   ],
   "source": [
    "from collections import Counter\n",
    "print (\"conteo de tweets clase Harassment\",Counter(df_train['harassment']))\n",
    "print (\"conteo de tweets clase IndirectH\",Counter(df_train['IndirectH']))\n",
    "print (\"conteo de tweets clase PhysicalH\",Counter(df_train['PhysicalH']))\n",
    "print (\"conteo de tweets clase SexualH\",Counter(df_train['SexualH']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 0, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0])"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_bi=df_train['harassment'].values\n",
    "y_train_bi[:20]\n",
    "#labels_mul= 0:non-harassment, 1:harassment, 2:IndirectH, 3:PhysicalH, 4:SexualH\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0. 1.]\n",
      " [1. 0.]\n",
      " [1. 0.]\n",
      " [0. 1.]\n",
      " [0. 1.]\n",
      " [1. 0.]\n",
      " [0. 1.]\n",
      " [0. 1.]\n",
      " [1. 0.]\n",
      " [1. 0.]]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(6374, 2)"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_bic=to_categorical(y_train_bi,num_classes=2)\n",
    "print (y_train_bic[0:10])\n",
    "y_train_bic.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "ys=df_train.values[:,2:]\n",
    "\n",
    "y_train_mul=np.zeros((df_train.shape[0],5))\n",
    "i=0\n",
    "for label in ys:\n",
    "    if np.sum(label)==0:\n",
    "        y_train_mul[i][0]=1\n",
    "    else: \n",
    "        y_train_mul[i]=np.concatenate([[0,0],label[1:]])\n",
    "    i+=1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Se encontraron 1193514 terminos con sus vectores de embedding.\n"
     ]
    }
   ],
   "source": [
    "data_train=[]\n",
    "for tw in df_train['tweet_content']:\n",
    "    data_train.append(cleaner(tw))\n",
    "\n",
    "    \n",
    "embeddings_index = {}\n",
    "#f = open(os.path.join(\"../../glove.twitter.27B/glove.twitter.27B.100d.txt\"))#'glove.twitter.27B.100d.txt'))\n",
    "f = open(os.path.join(\"../../Milenio/Twitter-stance-classification-by-TensorFlow/RNN_word2vec/data/glove.twitter.27B/glove.twitter.27B.100d.txt\")) \n",
    "for line in f: \n",
    "    values = line.split()\n",
    "    word = values[0]\n",
    "    coefs = np.asarray(values[1:], dtype='float32')\n",
    "    embeddings_index[word] = coefs\n",
    "f.close()\n",
    "\n",
    "print('Se encontraron %s terminos con sus vectores de embedding.' % len(embeddings_index))\n",
    "                                                                        \n",
    "word_index=dict()\n",
    "j=1\n",
    "for frase in data_train:\n",
    "    for term in frase:\n",
    "        if term not in word_index.keys():\n",
    "            word_index[term]=j\n",
    "            j+=1\n",
    "    \n",
    "embedding_vector=100\n",
    "embedding_matrix = np.zeros((len(word_index.keys())+1, embedding_vector))   #puede probar otra inicialización\n",
    "\n",
    "for word, i in word_index.items():\n",
    "    embedding_vector = embeddings_index.get(word)\n",
    "    if embedding_vector is not None:\n",
    "        embedding_matrix[i] = embedding_vector\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Counter({0: 3661, 1: 2713})\n",
      "[0.87052718 1.17471434]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f7b7e313240>"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b7e3076d8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print (Counter(y_train_bi))\n",
    "class_weights = class_weight.compute_class_weight('balanced', np.unique(y_train_bi), y_train_bi)#{0: 3.,   1: 6.,   2: 5.,  3: 3.}\n",
    "print (class_weights)\n",
    "x_new_train = [[word_index[word] for word in text] for text in data_train]\n",
    "\n",
    "sns.boxplot(list(map(len,x_new_train)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "76\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(6374, 76)"
      ]
     },
     "execution_count": 87,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "max_input_lenght = max(list(map(len,x_new_train))) #modificar este valor en base a su experimentación\n",
    "X_train = sequence.pad_sequences(x_new_train,maxlen=max_input_lenght,padding='post',value=0)\n",
    "print (max_input_lenght) \n",
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Carga data Validación"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "conteo de tweets clase Harassment Counter({0: 1493, 1: 632})\n",
      "conteo de tweets clase IndirectH Counter({0: 2054, 1: 71})\n",
      "conteo de tweets clase PhysicalH Counter({0: 2089, 1: 36})\n",
      "conteo de tweets clase SexualH Counter({0: 1600, 1: 525})\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1])"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print (\"conteo de tweets clase Harassment\",Counter(df_val['harassment']))\n",
    "print (\"conteo de tweets clase IndirectH\",Counter(df_val['IndirectH']))\n",
    "print (\"conteo de tweets clase PhysicalH\",Counter(df_val['PhysicalH']))\n",
    "print (\"conteo de tweets clase SexualH\",Counter(df_val['SexualH']))\n",
    "y_val_bin=df_val['harassment'].values\n",
    "y_val_bin[:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Counter({0: 1493, 1: 632})\n"
     ]
    }
   ],
   "source": [
    "ys_v=df_val.values[:,2:]\n",
    "\n",
    "y_val_mul=np.zeros((df_val.shape[0],5))\n",
    "i=0\n",
    "for label in ys_v:\n",
    "    if np.sum(label)==0:\n",
    "        y_val_mul[i][0]=1\n",
    "    else: \n",
    "        y_val_mul[i]=np.concatenate([[0,0],label[1:]])\n",
    "    i+=1\n",
    "\n",
    "data_val=[]\n",
    "for tw in df_val['tweet_content']:\n",
    "    data_val.append(cleaner(tw))\n",
    "print (Counter(y_val_bin))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2125, 76)\n",
      "[[1. 0.]\n",
      " [1. 0.]\n",
      " [0. 1.]]\n"
     ]
    }
   ],
   "source": [
    "Vec_train=[]\n",
    "tweets_c_a=data_train\n",
    "vocab_train=[]\n",
    "vocab=list(embeddings_index.keys())\n",
    "for post in tweets_c_a:\n",
    "    posteo=[]\n",
    "    for term in post:   #para todos los terminos de un tweet \n",
    "        vector=embeddings_index.get(term)\n",
    "        if vector is None:       \n",
    "            posteo.append(np.zeros(100))\n",
    "        else:\n",
    "            posteo.append(np.asarray(vector))\n",
    "            vocab_train.append(term)\n",
    "    Vec_train.append(np.asarray(posteo))\n",
    "    \n",
    "Vec_val_ig=[]\n",
    "tweets_l_vali=data_val\n",
    "vocab=list(embeddings_index.keys())\n",
    "for post in tweets_l_vali:\n",
    "    temp=[]\n",
    "    for term in post:  \n",
    "        if term in vocab_train:\n",
    "            vector=embeddings_index.get(term)\n",
    "            if vector is None:       \n",
    "                temp.append(np.zeros(100))\n",
    "            else:\n",
    "                temp.append(np.asarray(vector))\n",
    "    Vec_val_ig.append(temp)\n",
    "\n",
    "all_val=[]\n",
    "for text in data_val:\n",
    "    temp=[]\n",
    "    for word in text:\n",
    "        try:\n",
    "            temp.append(word_index[word])\n",
    "        except:\n",
    "            continue\n",
    "    all_val.append(temp)\n",
    "    \n",
    "x_new_val=all_val\n",
    "X_val_ig = sequence.pad_sequences(x_new_val,maxlen=max_input_lenght,padding='post',value=0)\n",
    "print (np.asarray(X_val_ig).shape)\n",
    "y_val_bic=to_categorical(y_val_bin,num_classes=2)\n",
    "print (y_val_bic[0:3])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "76"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "max_input_lenght"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import KFold\n",
    "from sklearn.metrics import accuracy_score,f1_score, confusion_matrix\n",
    "from sklearn.preprocessing import normalize\n",
    "import time\n",
    "\n",
    "def plot_confusion_matrix(cm, target_names, title='Confusion matrix (f1-score)',cmap=None, normalize=True):\n",
    "    \n",
    "    import itertools\n",
    "    accuracy = np.trace(cm) / float(np.sum(cm))\n",
    "    misclass = 1 - accuracy\n",
    "    if cmap is None:\n",
    "        cmap = plt.get_cmap('Blues')\n",
    "\n",
    "    plt.figure(figsize=(8, 6))\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "\n",
    "    if target_names is not None:\n",
    "        tick_marks = np.arange(len(target_names))\n",
    "        plt.xticks(tick_marks, target_names, rotation=45)\n",
    "        plt.yticks(tick_marks, target_names)\n",
    "\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "    thresh = cm.max() / 1.5 if normalize else cm.max() / 2\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        if normalize:\n",
    "            plt.text(j, i, \"{:0.4f}\".format(cm[i, j]),\n",
    "                     horizontalalignment=\"center\",\n",
    "                     color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "        else:\n",
    "            plt.text(j, i, \"{:,}\".format(cm[i, j]),\n",
    "                     horizontalalignment=\"center\",\n",
    "                     color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label\\naccuracy={:0.4f}; misclass={:0.4f}'.format(accuracy, misclass))\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Modelo Conv\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f7bcb929cf8>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaAAAAEECAYAAABuu+ChAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAF0lJREFUeJzt3X2wHXd93/H3BymOHyU/wKV4HJBdKpFMqC3bPHVIcJFpqDPQsakpD+axtSMYjEMZ1YVCA54JqWQCpQ5gWwMB8+TOxHaAFDB2y5SHAhYWNWCCVEs2TxlzjYKuJEtWuNa3f+ye5HCQdc+x7z170H2/ZnbOOb/v7t7fzmj0md/ub3dTVUiSNG6P6roDkqTFyQCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdWJp1x2YZI9+9KNrxYoVXXdDkn6l3H777T+tqsfMtZ4BdAgrVqzgG9/4RtfdkKRfKUm+P8x6noKTJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIbURfYWeuu67oLmkC3X/nyrrsgdc4RkCSpEwaQJKkTBpAkqRMGkCSpEwaQJKkTYw2gJH+c5O4ku5JMJ/mLJI9va69MciDJnr7lEwPbn53ktiR7k2xLctFAfSrJjUl2J7kvyfokhqwkTaBx/+f8EeCMqloGrAB+AFzfV99eVcf2LS/uFZIsBz4L3ACcAKwFrk7yjL7tP9Z+ngI8DTgfWLdQByNJevjGGkBV9b2qmml/BjgArBpy8wuAfcCGqtpfVbcANwGXACQ5FTgXWFdVM1W1HVhPE1RDS3JSkpVJVs7Ozo6yqSRpBGM/PZXkJUlmgD3AZcDb+sq/keTeJD9Mcn0bKj2nA5urqvraNrftvfpMVW0bqK9IsmyELl4KbAG2TE9Pj7CZJGkUYw+gqvp4VS0HHkcTPt9uS18EngycDDwFeAC4Jckxbf04YOYX98ZOYNkcdfrWGcZVNKOyVVNTUyNsJkkaRWcX6KvqXmAj8FdJTqyq7VW1taoOtLWLacLo6e0mu4HlA7s5Htg1R71XG7ZfO9p+bF261CcVSdJC6XqG2FLgGJqgGVTtkvb3HcDqgXVWt+29+vIkpw3U7+m77iRJmhBjC6Akj0ryuiRT7e9TgPcC9wDfS/L7SU5J48S29lPga+0ubgKOTrIuyRFJ1tBMTLgWoKruBm4FNiRZ1l4/uhy4ZlzHKEka3rhHQOcB30lyP/B1YC9wblXNAucAt9FMTrgTOAl4TlXtAaiqne32F9Jc69kIrK2qr/bt/6U0x/RjYBPwSWDDwh+WJGlUY7vIUVUHaALkoerrmOOenaraBDz1EPVpmlGRJGnCdX0NSJK0SBlAkqROGECSpE4YQJKkThhAkqROGECSpE4YQJKkThhAkqROGECSpE4YQJKkThhAkqROGECSpE4YQJKkThhAkqROGECSpE4YQJKkThhAkqROGECSpE6MNYCS/HGSu5PsSjKd5C+SPL6v/vIk25LsTfL1JGcNbH92ktva+rYkFw3Up5LcmGR3kvuSrE9iyErSBBr3f84fAc6oqmXACuAHwPUASZ4JvB94DXACcAPwmSTL2vpy4LNt+wnAWuDqJM/o2//H2s9TgKcB5wPrFvaQJEkPx1gDqKq+V1Uz7c8AB4BV7e+LgRur6vNVtR+4EthPEyIAFwD7gA1Vtb+qbgFuAi4BSHIqcC6wrqpmqmo7sJ4mqCRJE2bsp6eSvCTJDLAHuAx4W1s6Hbi9t15VFfDNtr1X39y292weqM9U1baB+oreKGrI/p2UZGWSlbOzs8MfmCRpJGMPoKr6eFUtBx5HEz7fbkvHATMDq+8Elj3COn3rDONSYAuwZXp6eoTNJEmj6OwCfVXdC2wE/irJicBuYPnAascDu9rvD7feqw3rKprTgqumpqZG2EySNIquZ4gtBY4BTgbuAM7sFZIEOKNtp/1cPbD96oH68iSnDdTv6bvuNKeq2lFVW6tq69KlS0c5FknSCMYWQEkeleR1Saba36cA7wXuAb5HMxq6IMmaJEcAbwSOpJloQPt5dJJ1SY5IsoZmYsK1AFV1N3ArsCHJsnZSwuXANeM6RknS8MY9AjoP+E6S+4GvA3uBc6tqtqq+DLyWJohmgBcC51XVLoCq2tluf2Fb3wisraqv9u3/pTTH9GNgE/BJYMM4DkySNJqxnWOqqgM0AXKoda4DrjtEfRPw1EPUp2lGRZKkCdf1NSBJ0iJlAEmSOmEASZI6YQBJkjphAEmSOmEASZI6YQBJkjphAEmSOmEASZI6YQBJkjrh456lReoHVzy56y5oAj3+P3977pXmiSMgSVInDCBJUicMIElSJwwgSVInDCBJUicMIElSJwwgSVInDCBJUifGFkBJ1ie5M8muJH+TZGOSE/vqr0xyIMmevuUTA/s4O8ltSfYm2ZbkooH6VJIbk+xOcl/7Nw1ZSZpA4/zP+UHgIuAk4HTgFODPB9bZXlXH9i0v7hWSLAc+C9wAnACsBa5O8oy+7T/Wfp4CPA04H1i3EAcjSXpkxhZAVfXmqvpmVf28qu4D/gw4Z4RdXADsAzZU1f6qugW4CbgEIMmpwLnAuqqaqartwHqaoBpakpOSrEyycnZ2dpRNJUkj6PL01BrgWwNtv5Hk3iQ/THJ9Gyo9pwObq6r62ja37b36TFVtG6ivSLJshH5dCmwBtkxPT4+wmSRpFJ0EUJIXABcDl/U1fxF4MnAy8BTgAeCWJMe09eOAmYFd7QSWzVGnb51hXAWsAlZNTU2NsJkkaRRjD6AkFwIbgedX1eZee1Vtr6qtVXWgqu6lCaiTgae3q+wGlg/s7nhg1xz1Xm0oVbWj7cfWpUt9WLgkLZSxBlCSVwHXAM+rqi/MsXq1S9rfdwCrB9ZZ3bb36suTnDZQv6eqBkdGkqSOjXMa9uuBdwK/V1VfOUj995OcksaJwHuBnwJfa1e5CTg6ybokRyRZQzMx4VqAqrobuBXYkGRZe/3ocprAkyRNmHGOgN5Dcy3mC/33+vTVzwFuA/YAd9JM135OVe0BqKqdwHnAhTTXejYCa6vqq337eCnNMf0Y2AR8EtiwkAclSXp4xnaRo6oyR30dc9yzU1WbgKceoj5NMyqSJE04nxIgSeqEASRJ6oQBJEnqhAEkSeqEASRJ6oQBJEnqhAEkSeqEASRJ6oQBJEnqxNABlOTBJL/0foL2BW4Pzm+3JEmHu1FGQA/1KJ0jgJ/PQ18kSYvInM+CS/Ly9msBL0yyq6+8BHgWcNcC9E2SdBgb5mGkH+j7/u6B2t8B24E3zFuPJEmLwpwBVFW/BpDkbuApVfXTBe+VJOmwN/TrGKrq1IXsiCRpcRnpfUDtW0bPAR7LwASGqnrH/HVLknS4GzqAkrwU+HOaGW/30UxK6CnAAJIkDW2UEdDbgf8KvLmqZheoP5KkRWKU+4AeB1z9cMMnyfokdybZleRvkmxMcuLAOi9Psi3J3iRfT3LWQP3sJLe19W1JLhqoTyW5McnuJPe1f9OnPUjSBBrlP+cvAGc8gr/1IHARcBJwOnAKzSk9AJI8E3g/8BrgBOAG4DNJlrX15cBn2/YTgLXA1Ume0fc3PtZ+ngI8DTgfWPcI+ixJWiCjBNB1wIYklyV5VpJ/1r/MtXFVvbmqvllVP6+q+4A/o5nQ0HMxcGNVfb6q9gNXAvtpQgTgAmAfsKGq9lfVLcBNwCXw9xMkzgXWVdVMVW0H1tME1dDaRwutTLJydtYzjZK0UEa5BnR9+zl4Myo0kxCWjPi31wDf6vt9OvChv99hVSX5Ztveq2+uqv7JD5uBl/XVZ6pq20B9RZJlVdX/BIdDuRT4I4Dp6ekhN5EkjWqUAJq3+4CSvIBmxPOsvubjgJmBVXcCyx5hnXadYQPoKuDjAFNTU1uG3EaSNKJRbkT9/nz8wSQXAtcAz6+qzX2l3cDygdWPB7b11VccpL6rr36w7Xu1oVTVDmAHwNlnnz3sZpKkEY1yH9BLDlWvqo8PsY9XAX8KPK+qvjJQvgM4s2/d0Ex6uLGvfv7ANqvb9l59eZLT2us/vfo9VTU4MpIkdWyUU3AffYj23jWZQwZQktfTXFv5varadJBVNgKfS/Jh4EvA64EjaSYa0H5uSLIOeA/wOzQTE54DUFV3J7m1XefVNLPtLqcZbUmSJszQs+Cq6lH9C817gJ4OfAV45hC7eA/NtZgvJNnTW/r2/2XgtTRBNAO8EDivN3mgqnYC5wEXtvWNwNqq+mrf33hpe0w/BjYBnwQ2DHuMkqTxGelZcP3aG1JvS/KfgPfSd/rsIdZ/qBfa9a9zHc1074eqbwKeeoj6NM2oSJI04ebjKQHTwKp52I8kaREZZRLCyYNNwMnA24DvzmOfJEmLwCin4H7ELz4BG5oQ+j7N9RpJkoY2SgD984HfB2hOv91VVQ/OX5ckSYvBKDei/u+F7IgkaXEZ9Y2ojwFeB/xW2/Qd4H3tw0UlSRra0LPgkjwduAt4VV/zq4H/l+Qhp0ZLknQwo4yA3gn8JfBvey+lS7IE+ADN43V+Z/67J0k6XI0SQGcBF/e/EbWqHkyyAfjGvPdMknRYG+VG1D3APzpI+xRw//x0R5K0WIwSQJ8CNiY5N8lR7bIGuJrm1JwkSUMb5RTcG2jeWPp5fvGG1JuAN85jnyRJi8Ao9wHtAi5I8kTgN9vm7w68AluSpKGM8iy4DwLfqap30UzH7rX/e+C3qurfLUD/JEmHqVGuAT0X+J8Haf8C8C/npzuSpMVilAA6kWYm3KBdNG8flSRpaKME0Hbg2QdpfzbNE7ElSRraKLPgrgbemeRo4Na2bQ1wBfD2+e6YJOnwNvQIqKr+G3AV8CfAt9plPfDeqnr3MPtI8qIkX0qyK8nsQO2cJJVkT9/yfwbWeWKSW5Pcn+RHSd44UD86yQeT/CzJziQfSHLUsMcoSRqfkZ6GXVVvSfIn/MPTsL9bVaM8BeFnwPuAo4BrD1J/sKqOPdiG7XPnPk0z+no+8CTgc0l+VFX/vV3tPW37k2juVfpL4F3Aa0booyRpDEa5BgRAVd1fVZvaZaRH8FTVzVX1CZrrSaP6XeAJwJuqam9VbQauAdYCtCOdi4C3VtVPqmoaeCvwiiRHDvtHkpyUZGWSlbOzs3NvIEl6WEYOoAW2JMkPk9yb5H8kOb2vdjqwtar6Z+JtbtsBVgFHArcP1I8CVo7Qh0uBLcCW6enpkQ9AkjScSQqg7wFnAKfSnEL7FvC/kpzc1o8DZga22Qks66szsE7v+zKGdxVNmK2ampoaYTNJ0igmJoCq6t6quqOqZqtqZ1W9Cfhb/uEm193A8oHNjqe5D6lXZ2Cd3vddDKmqdlTV1qraunTpSJfIJEkjmJgAeggHgLTf7wBWJjmmr766bYfmtNkDwJkD9X3A1gXupyRpRGMNoCRL2gkBR7S/j2yXJHl2O836UUmOTfI24LHAze3mX6S54fUd7asgzgD+gGYiAlW1D/gocEWSqSRTNPcoXVdVD4zzOCVJcxv3COhlNCOSm4El7fd9NLPbTqd51txumllyTweeU1U/hObtq8DzgN8GdgCfAa6squv79n8ZzWint2yheY2EJGnCjPUiR1V9iOadQgfz7nY51PZ30Tx94aHqe4FXt4skaYJN+jUgSdJhygCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIAkiR1YqwBlORFSb6UZFeS2YPUn5vkziT7knwnyb8YqD8xya1J7k/yoyRvHKgfneSDSX6WZGeSDyQ5aqGPS5I0unGPgH4GvA/4w8FCktOAG4E/AZa3nzclWdHWlwCfBv4aeAzwfODyJP+mbzfvAZ7ULiuB3wTetTCHIkl6JMYaQFV1c1V9Ath+kPIrgNur6qNV9XdV9TFgc9sO8LvAE4A3VdXeqtoMXAOsBWhHOhcBb62qn1TVNPBW4BVJjhy2j0lOSrIyycrZ2V8apEmS5skkXQM6Hbh9oG1z296rb62qPQ9RXwUcObCPzcBRNKOhYV0KbAG2TE9Pj7CZJGkUkxRAxwEzA207gWUj1BlYp/d9GcO7iibMVk1NTY2wmSRpFJMUQLtprv30Ox7YNUKdgXV633cxpKraUVVbq2rr0qVLh91MkjSiSQqgO4AzB9pWt+29+sokxzxEfQvwwMA+VgP7gK3z3ltJ0iMy7mnYS9oJAUe0v49slwDXAWcneXGSX0vyYuAs4MPt5l8Evg+8I8lRSc4A/oBmIgJVtQ/4KHBFkqkkU8AVwHVV9cA4j1OSNLdxj4BeRjMiuRlY0n7fBzyhqrYBFwBvoTll9hbg/Kq6B6CqHgSeB/w2sAP4DHBlVV3ft//LaEY7vWUL8IYFPypJ0sjGepGjqj4EfOgQ9c8BnztE/S5gzSHqe4FXt4skaYJN0jUgSdIiYgBJkjphAEmSOmEASZI6YQBJkjphAEmSOmEASZI6YQBJkjphAEmSOmEASZI6YQBJkjphAEmSOmEASZI6YQBJkjphAEmSOmEASZI6YQBJkjphAEmSOjFRAZTkQ0l+nmRP3/LagXVenmRbkr1Jvp7krIH62Ulua+vbklw03qOQJA1jogKo9eGqOrZveV+vkOSZwPuB1wAnADcAn0myrK0vBz7btp8ArAWuTvKMcR+EJOnQJjGADuVi4Maq+nxV7QeuBPYD57f1C4B9wIaq2l9VtwA3AZcM+weSnJRkZZKVs7Oz89x9SVLPJAbQC5L8bZKtSa5Mcmxf7XTg9t6Pqirgm217r765be/Z3FcfxqXAFmDL9PT0wzoASdLcJi2ArgKeBDyaZlTzLGBjX/04YGZgm53AsiHrw/ZhFbBqampqhM0kSaOYqACqqtur6idVdaCq7gTeAPzrJL/errIbWD6w2fHAriHrw/RhR1VtraqtS5cuHf0gJElDmagAOogD7WfazzuAM3vFJAHOaNt79dUD+1jdV5ckTYiJCqAkL0pyfPv9nwB/Cnyqqh5oV9kIXJBkTZIjgDcCR9JMNKD9PDrJuiRHJFlDMzHh2rEeiCRpThMVQDTTprcnuR/4PPA14FW9YlV9GXgtTRDNAC8EzquqXW19J3AecGFb3wisraqvjvMgJElzm6iLHFV1zhDrXAdcd4j6JuCp89gtSdICmLQRkCRpkTCAJEmdMIAkSZ0wgCRJnTCAJEmdMIAkSZ0wgCRJnTCAJEmdMIAkSZ0wgCRJnTCAJEmdMIAkSZ0wgCRJnTCAJEmdMIAkSZ0wgCRJnTCAJEmdMIAkSZ047AIoyZIkVya5L8nuJDckeXTX/ZIk/aLDLoCA/wj8K+BpwClt20e6644k6WAOxwC6BFhfVduragb4D8Bzk6wYZuMkJyVZmWTl7OzsAnZTkha3VFXXfZg3SZYDO4HVVfV/+9pngJdV1aeG2MfbgD9qf+4F/noBuroYLQEeC/wEeLDjvkj9/Lc5/55QVY+Za6Wl4+jJGC1rP2cG2nf21eZyFfDx9vuOqtoxHx1b7JKsBLYA51TV1q77I/X4b7M7h1sA7W4/lw+0Hw/sGmYHbeAYOpK0wA6ra0BVtRP4AXBmry3JaTSjn2911S9J0i87rAKodS1weZJTkywD1gM3V9U93XZr0dsBvB1Hl5o8/tvsyGE1CQGa+4BoQueVwK8DtwCXVNVPu+yXJOkXHXYBJEn61XA4noKTJP0KMIAkSZ0wgCRJnTCAJEmdMIAkSZ0wgCRJnTCAJEmdMIAkSZ0wgLSgfEOtJlWSFyX5UpJdSXz5VwcMIC0031CrSfUz4H3AH3bdkcXKR/FoQSX5PnBFVX2g/f2PgbuAU31ArCZBknOAW6vqcHs9zcRzBKQF076h9vHA7b22qtpG826mf9pVvyRNBgNIC2k+3lAr6TBlAGkhPeI31Eo6fBlAWjC+oVbSoRhAWmi+oVYTqb1F4EjgiPb3ke2Sjru2aBhAWmj/Bfg0sAn4MbAEuKjTHkmNlwH7gJtp/l3ua5cndNmpxcRp2JKkTjgCkiR1wgCSJHXCAJIkdcIAkiR1wgCSJHXCAJIkdcIAkiR1wgCSJHXi/wMm8s7ZYmoCNgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7bcb930470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.countplot(y_train_bi)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from keras import backend as K\n",
    "\n",
    "def focal_loss(gamma=2., weights=1):   #weights np.asarray()\n",
    "    weights= K.variable(weights)\n",
    "    def focal_loss_fixed(y_true, y_pred):\n",
    "        y_true = K.clip(y_true, K.epsilon(),1)\n",
    "        y_pred = K.clip(y_pred,K.epsilon(),1)\n",
    "        return - K.sum(weights* K.pow(1. - y_pred, gamma)* y_true * K.log(y_pred), axis=-1) \n",
    "    return focal_loss_fixed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "embedding_vector=100\n",
    "embedding_layer = Embedding(input_dim=len(word_index.keys())+1,output_dim=embedding_vector,weights=[embedding_matrix],\n",
    "                     input_length=max_input_lenght,trainable=False)\n",
    "\n",
    "sequence_input = Input(shape=(max_input_lenght,))\n",
    "embedded_sequences = embedding_layer(sequence_input)\n",
    "batch = BatchNormalization()(embedded_sequences)\n",
    "cov1= Conv1D(128, 5, activation='relu',padding='same')(batch)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)    \n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "cov1= Conv1D(64, 5, activation='relu',padding='same')(drop)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)\n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "flat = Flatten()(drop)\n",
    "preds = Dense(64, activation='relu')(flat)\n",
    "preds = BatchNormalization()(preds)\n",
    "preds = Dense(1, activation='sigmoid')(preds)\n",
    "model = Model(sequence_input, preds)\n",
    "model.compile(loss='binary_crossentropy',optimizer='adam',metrics=['acc'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "history = model.fit(X_train, y_train_bi,batch_size=64,epochs=10,verbose=0,validation_data=(X_val_ig,y_val_bin),class_weight=class_weights)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2125, 76)"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_val_ig.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.save('matrices/X_train_binario.npy',X_train)\n",
    "np.save('matrices/y_train_bi_binario.npy',y_train_bi)\n",
    "np.save('matrices/X_val_ig_binario.npy',X_val_ig)\n",
    "np.save('matrices/y_val_bin_binario.npy',y_val_bin)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.03730239]\n",
      " [0.06119494]\n",
      " [0.9617095 ]\n",
      " [0.05833904]\n",
      " [0.6518054 ]]\n",
      "[array([0.], dtype=float32), array([0.], dtype=float32), array([1.], dtype=float32), array([0.], dtype=float32), array([1.], dtype=float32)]\n",
      "Accuracy sobre validacion: 0.900235294117647\n",
      "F1-score None sobre validación: [0.93089961 0.82064298]\n",
      "F1-score macro sobre validación: 0.8757712934345474\n",
      "Precision None sobre validación: [0.90666667 0.88181818]\n",
      "Recall None sobre validación: [0.9564635  0.76740506]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b9e848ba8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "trainPredict = model.predict(X_val_ig, batch_size=64)\n",
    "print (trainPredict[:5])\n",
    "trainPredict=[np.round(pred) for pred in trainPredict]\n",
    "print (trainPredict[:5])\n",
    "acc_val= accuracy_score(y_val_bin, trainPredict)\n",
    "f1_val=f1_score(y_val_bin, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_bin, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_bin, trainPredict, average=None) \n",
    "pre=precision_score(y_val_bin, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_bin, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)\n",
    "\n",
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),[\"NonH\",\"H\"],\"Confusion Matrix Caso Binario 100dim\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model.save('baselines/cnn1_binary_ce10_100.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy sobre validacion: 0.8936470588235295\n",
      "F1-score None sobre validación: [0.92742453 0.80105634]\n",
      "F1-score macro sobre validación: 0.8642404361945597\n",
      "Precision None sobre validación: [0.89080814 0.90277778]\n",
      "Recall None sobre validación: [0.96718017 0.71993671]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b9c9f0f28>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "embedding_vector=100\n",
    "embedding_layer = Embedding(input_dim=len(word_index.keys())+1,output_dim=embedding_vector,weights=[embedding_matrix],\n",
    "                     input_length=max_input_lenght,trainable=False)\n",
    "\n",
    "sequence_input = Input(shape=(max_input_lenght,))\n",
    "embedded_sequences = embedding_layer(sequence_input)\n",
    "batch = BatchNormalization()(embedded_sequences)\n",
    "cov1= Conv1D(128, 5, activation='relu',padding='same')(batch)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)    \n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "cov1= Conv1D(64, 5, activation='relu',padding='same')(drop)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)\n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "flat = Flatten()(drop)\n",
    "preds = Dense(64, activation='relu')(flat)\n",
    "preds = BatchNormalization()(preds)\n",
    "preds = Dense(2, activation='softmax')(preds)\n",
    "model = Model(sequence_input, preds)\n",
    "model.compile(loss=focal_loss(2),optimizer='adam',metrics=['acc'])\n",
    "\n",
    "history = model.fit(X_train, y_train_bic,batch_size=64,epochs=10,verbose=0,validation_data=(X_val_ig,y_val_bic),class_weight=class_weights)\n",
    "trainPredict = model.predict(X_val_ig, batch_size=64).argmax(axis=-1)\n",
    "acc_val= accuracy_score(y_val_bin, trainPredict)\n",
    "f1_val=f1_score(y_val_bin, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_bin, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_bin, trainPredict, average=None) \n",
    "pre=precision_score(y_val_bin, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_bin, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)\n",
    "\n",
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),[\"NonH\",\"H\"],\"Confusion Matrix Caso Binario + Focal Loss 100dim\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model.save('baselines/cnn1_binary_focal10_100.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy sobre validacion: 0.8927058823529411\n",
      "F1-score None sobre validación: [0.92682927 0.7989418 ]\n",
      "F1-score macro sobre validación: 0.862885533617241\n",
      "Precision None sobre validación: [0.88971041 0.90239044]\n",
      "Recall None sobre validación: [0.96718017 0.71677215]\n"
     ]
    }
   ],
   "source": [
    "# lo mismo con focal + class\n",
    "embedding_vector=100\n",
    "embedding_layer = Embedding(input_dim=len(word_index.keys())+1,output_dim=embedding_vector,weights=[embedding_matrix],\n",
    "                     input_length=max_input_lenght,trainable=False)\n",
    "\n",
    "sequence_input = Input(shape=(max_input_lenght,))\n",
    "embedded_sequences = embedding_layer(sequence_input)\n",
    "batch = BatchNormalization()(embedded_sequences)\n",
    "cov1= Conv1D(128, 5, activation='relu',padding='same')(batch)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)    \n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "cov1= Conv1D(64, 5, activation='relu',padding='same')(drop)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)\n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "flat = Flatten()(drop)\n",
    "preds = Dense(64, activation='relu')(flat)\n",
    "preds = BatchNormalization()(preds)\n",
    "preds = Dense(2, activation='softmax')(preds)\n",
    "model = Model(sequence_input, preds)\n",
    "model.compile(loss=focal_loss(2, class_weights),optimizer='adam',metrics=['acc'])\n",
    "\n",
    "history = model.fit(X_train, y_train_bic,batch_size=64,epochs=10,verbose=0,validation_data=(X_val_ig,y_val_bic),class_weight=class_weights)\n",
    "\n",
    "trainPredict = model.predict(X_val_ig, batch_size=64).argmax(axis=-1)\n",
    "acc_val= accuracy_score(y_val_bin, trainPredict)\n",
    "f1_val=f1_score(y_val_bin, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_bin, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_bin, trainPredict, average=None) \n",
    "pre=precision_score(y_val_bin, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_bin, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b9c282208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),[\"NonH\",\"H\"],\"Confusion Matrix Caso Binario + Focal 100dim\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "## sin guardar por no ser competitivo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Caso multiclass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2713\n",
      "632\n",
      "[['pic', 'mexican', 'school', 'girl', 'naked', 'melanie', 'safka', 'fake', 'nude', 'photo', 'hardcore', 'slut', 'movie', 'alicia', 'moore', 'porn', 'star', 'toon', 'love'], ['girl', 'dirty', 'quarterback', 'latino', 'pokemon', 'young', 'handjob', 'kissing', 'gangbang', 'w', 'x89', 't', 'co', 'wrm4xo00vj'], ['miliondollameat', 'sexting', 'n', 'ngirl', 'daddy', 'n', 'nme', 'listen', 'bitch', 'gonna', 'watch', 'these', 'what'], ['museum', 'slut', 'if', 'you', 'know', 'somewhere', 'we', 'can', 'do', 'it', 'then', 'i', 'am', 'game'], ['rt', 'sp00kybitchhhh', 'i', 'have', 'spot', 'open', 'sissy', 'slut', 'intox', 'bitch', 'human', 'atm', 'amazon', 'boy', 'do', 'you', 'have', 'what', 'it', 'take', 'can', 'you', 'live', 'up', 'to', 'my']]\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "7615"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unnamed_har=np.where(df_train['harassment']== 1)[0]\n",
    "print (len(unnamed_har))\n",
    "unnamed_har_val=np.where(df_val['harassment']== 1)[0]\n",
    "print (len(unnamed_har_val))\n",
    "\n",
    "df_train_f2=pd.DataFrame(df_train.values[unnamed_har], columns=['unnamed_har', 'tweet','harassment','IndirectH','PhysicalH','SexualH'])\n",
    "df_val_f2=pd.DataFrame(df_val.values[unnamed_har_val], columns=['unnamed_har', 'tweet','harassment','IndirectH','PhysicalH','SexualH'])\n",
    "\n",
    "data_train_f2=[]\n",
    "for tw in df_train_f2['tweet']:\n",
    "    data_train_f2.append(cleaner(tw))\n",
    "print (data_train_f2[:5])\n",
    "\n",
    "word_index_f2=dict()\n",
    "j=1\n",
    "for frase in data_train_f2:\n",
    "    for term in frase:\n",
    "        if term not in word_index_f2.keys():\n",
    "            word_index_f2[term]=j\n",
    "            j+=1\n",
    "    \n",
    "len(word_index_f2.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[16.44242424 11.89912281  0.35024529]\n"
     ]
    }
   ],
   "source": [
    "embedding_vector=100\n",
    "embedding_matrix_f2 = np.zeros((len(word_index_f2.keys())+1, embedding_vector))   #puede probar otra inicialización\n",
    "\n",
    "for word, i in word_index_f2.items():\n",
    "    embedding_vector = embeddings_index.get(word)\n",
    "    if embedding_vector is not None:\n",
    "        embedding_matrix_f2[i] = embedding_vector\n",
    "y_train_f2=np.asarray(np.asarray(df_train_f2.values)[:,-3:],dtype='float32')\n",
    "y_train_mu=[] #123\n",
    "for array in y_train_f2:\n",
    "    y_train_mu.append(np.argmax(array))\n",
    "y_train_mul=to_categorical(y_train_mu,num_classes=3)\n",
    "\n",
    "class_weights_f2 = class_weight.compute_class_weight('balanced', np.unique(y_train_mu), y_train_mu)\n",
    "print (class_weights_f2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(0, 55), (1, 76), (2, 2582)]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b6d4b75c0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "c = Counter(y_train_mu)\n",
    "c = sorted(c.items())\n",
    "print (c)\n",
    "clases=['Indirect','Physical','Sexual']\n",
    "class_num = [i[0] for i in c]\n",
    "class_names = [clases[i[0]-1] for i in c]\n",
    "freq = [i[1] for i in c]\n",
    "#plt.bar(y_train_mu)\n",
    "\n",
    "f, ax = plt.subplots(figsize=(8,5))\n",
    "plt.bar(class_num, freq)\n",
    "plt.title(\"# ataques por clase de acoso\")\n",
    "plt.xlabel(\"Tipo de acoso\")\n",
    "plt.ylabel(\"Frequencia\")\n",
    "ax.set_xticks(range(0, 3))\n",
    "ax.set_xticklabels(clases)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f7b6d4fc8d0>"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b6d4fc588>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x_new_train = [[word_index_f2[word] for word in cleaner(text)] for text in df_train_f2[\"tweet\"].values]\n",
    "sns.boxplot(list(map(len,x_new_train)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2713\n",
      "632\n"
     ]
    }
   ],
   "source": [
    "max_input_lenght = max(list(map(len,x_new_train))) #modificar este valor en base a su experimentación\n",
    "X_train_f2 = sequence.pad_sequences(x_new_train,maxlen=max_input_lenght,padding='post',value=0)\n",
    "\n",
    "list_posts=[cleaner(text) for text in df_train_f2[\"tweet\"].values]  #train\n",
    "print (len(list_posts))\n",
    "\n",
    "Vec_train=[]\n",
    "tweets_c_a=list_posts\n",
    "vocab_train_f2=[]\n",
    "vocab=list(embeddings_index.keys())\n",
    "for post in list_posts:\n",
    "    posteo=[]\n",
    "    for term in post:   #para todos los terminos de un tweet \n",
    "        vector=embeddings_index.get(term)\n",
    "        if vector is None:       \n",
    "            posteo.append(np.zeros(100))\n",
    "        else:\n",
    "            posteo.append(np.asarray(vector))\n",
    "            vocab_train_f2.append(term)\n",
    "    Vec_train.append(np.asarray(posteo))\n",
    "    \n",
    "list_posts_val=[cleaner(text) for text in df_val_f2[\"tweet\"].values]  #train\n",
    "print (len(list_posts_val))\n",
    "\n",
    "Vec_val_ig=[]\n",
    "vocab=list(embeddings_index.keys())\n",
    "for post in list_posts_val:\n",
    "    temp=[]\n",
    "    for term in post:   #para todos los terminos de un tweet \n",
    "        if term in vocab_train_f2:\n",
    "            vector=embeddings_index.get(term)\n",
    "            if vector is None:       \n",
    "                temp.append(np.zeros(100))\n",
    "            else:\n",
    "                temp.append(np.asarray(vector))\n",
    "    Vec_val_ig.append(temp)\n",
    "    \n",
    "word_index_f2_val=dict()\n",
    "j=1\n",
    "for frase in list_posts_val:\n",
    "    for term in frase:\n",
    "        if term not in word_index_f2_val.keys():\n",
    "            word_index_f2_val[term]=j\n",
    "            j+=1\n",
    "            \n",
    "all_val_dif=[]\n",
    "for text in list_posts_val:\n",
    "    temp=[]\n",
    "    for word in text:\n",
    "        try:\n",
    "            temp.append(word_index_f2_val[word])\n",
    "        except:\n",
    "            continue\n",
    "    all_val_dif.append(temp)\n",
    "\n",
    "all_val=[]\n",
    "for text in list_posts_val:\n",
    "    temp=[]\n",
    "    for word in text:\n",
    "        try:\n",
    "            temp.append(word_index_f2[word])\n",
    "        except:\n",
    "            continue\n",
    "    all_val.append(temp)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "32"
      ]
     },
     "execution_count": 99,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "max_input_lenght"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(632, 3)"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_new_val=all_val#.copy()\n",
    "X_val_ig = sequence.pad_sequences(x_new_val,maxlen=max_input_lenght,padding='post',value=0)\n",
    "y_val_f2=np.asarray(np.asarray(df_val_f2.values)[:,-3:],dtype='float32')\n",
    "y_val_mu=[] #123\n",
    "for array in y_val_f2:\n",
    "    y_val_mu.append(np.argmax(array))\n",
    "y_val_mul=to_categorical(y_val_mu,num_classes=3)\n",
    "y_val_mul.shape#one hot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(0, 71), (1, 36), (2, 525)]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b6dcc9a90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "c = Counter(y_val_mu)\n",
    "c = sorted(c.items())\n",
    "print (c)\n",
    "clases=['Indirect','Physical','Sexual']\n",
    "class_num = [i[0] for i in c]\n",
    "class_names = [clases[i[0]-1] for i in c]\n",
    "freq = [i[1] for i in c]\n",
    "#plt.bar(y_train_mu)\n",
    "\n",
    "f, ax = plt.subplots(figsize=(8,5))\n",
    "plt.bar(class_num, freq)\n",
    "plt.title(\"# ataques por clase de acoso VAL SET\")\n",
    "plt.xlabel(\"Tipo de acoso\")\n",
    "plt.ylabel(\"Frequencia\")\n",
    "ax.set_xticks(range(0, 3))\n",
    "ax.set_xticklabels(clases)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Conv 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy sobre validacion: 0.7753164556962026\n",
      "F1-score None sobre validación: [0.38167939 0.19565217 0.87608069]\n",
      "F1-score macro sobre validación: 0.4844707516228906\n",
      "Precision None sobre validación: [0.41666667 0.16071429 0.88372093]\n",
      "Recall None sobre validación: [0.35211268 0.25       0.86857143]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b948b9b70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "embedding_vector=100\n",
    "embedding_layer = Embedding(input_dim=len(word_index_f2.keys())+1,output_dim=embedding_vector,weights=[embedding_matrix_f2],\n",
    "                     input_length=max_input_lenght,trainable=False)\n",
    "\n",
    "sequence_input = Input(shape=(max_input_lenght,))\n",
    "embedded_sequences = embedding_layer(sequence_input)\n",
    "batch = BatchNormalization()(embedded_sequences)\n",
    "cov1= Conv1D(128, 5, activation='relu',padding='same')(batch)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)    \n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "cov1= Conv1D(64, 5, activation='relu',padding='same')(drop)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)\n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "flat = Flatten()(drop)\n",
    "preds = Dense(100, activation='relu')(flat)\n",
    "preds = BatchNormalization()(preds)\n",
    "preds = Dense(3, activation='softmax')(preds)\n",
    "model = Model(sequence_input, preds)\n",
    "model.compile(loss=focal_loss(2,weights=class_weights_f2),optimizer='adam',metrics=['acc'])\n",
    "history = model.fit(X_train_f2, y_train_mul,batch_size=64,epochs=20,verbose=0,validation_data=(X_val_ig,y_val_mul),class_weight=class_weights_f2)   \n",
    "\n",
    "trainPredict = model.predict(X_val_ig, batch_size=64).argmax(axis=-1)\n",
    "acc_val= accuracy_score(y_val_mu, trainPredict)\n",
    "f1_val=f1_score(y_val_mu, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_mu, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_mu, trainPredict, average=None) \n",
    "pre=precision_score(y_val_mu, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_mu, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)\n",
    "\n",
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),['Indirect','Physical','Sexual'],\"Confusion Matrix Caso Multiclass + Focal Loss 100dim\")\n",
    "model.save('baselines/cnn1_100_multi_focal20_100.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy sobre validacion: 0.7278481012658228\n",
      "F1-score None sobre validación: [0.38235294 0.16528926 0.84210526]\n",
      "F1-score macro sobre validación: 0.46324915351090407\n",
      "Precision None sobre validación: [0.4        0.11764706 0.87966805]\n",
      "Recall None sobre validación: [0.36619718 0.27777778 0.80761905]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b90a2bba8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "embedding_vector=100\n",
    "embedding_layer = Embedding(input_dim=len(word_index_f2.keys())+1,output_dim=embedding_vector,weights=[embedding_matrix_f2],\n",
    "                     input_length=max_input_lenght,trainable=False)\n",
    "\n",
    "sequence_input = Input(shape=(max_input_lenght,))\n",
    "embedded_sequences = embedding_layer(sequence_input)\n",
    "\n",
    "batch = BatchNormalization()(embedded_sequences)\n",
    "cov1= Conv1D(128, 5, activation='relu',padding='same')(batch)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)    \n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "cov1= Conv1D(64, 5, activation='relu',padding='same')(drop)\n",
    "pool = MaxPooling1D(pool_size=3)(cov1)\n",
    "batch = BatchNormalization()(pool)\n",
    "drop = Dropout(0.65)(batch)\n",
    "flat = Flatten()(drop)\n",
    "preds = Dense(64, activation='relu')(flat)\n",
    "preds = BatchNormalization()(preds)\n",
    "preds= Dropout(0.3)(preds)\n",
    "preds = Dense(3, activation='softmax')(preds)\n",
    "model2 = Model(sequence_input, preds)\n",
    "model2.compile(loss=focal_loss(2,weights=class_weights_f2),optimizer='adam',metrics=['acc'])\n",
    "\n",
    "history = model2.fit(X_train_f2, y_train_mul,batch_size=64,epochs=20,verbose=0,validation_data=(X_val_ig,y_val_mul),class_weight=class_weights_f2)   \n",
    "\n",
    "trainPredict = model2.predict(X_val_ig, batch_size=64).argmax(axis=-1)\n",
    "acc_val= accuracy_score(y_val_mu, trainPredict)\n",
    "f1_val=f1_score(y_val_mu, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_mu, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_mu, trainPredict, average=None) \n",
    "pre=precision_score(y_val_mu, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_mu, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)\n",
    "\n",
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),['Indirect','Physical','Sexual'],\"Confusion Matrix Caso Multiclass + Focal Loss 100dim\")\n",
    "#model2.save('baselines/cnn2_128_multi_focal20_100.h5')  #[0.37179487 0.19753086 0.74856487]\n",
    "#F1-score macro sobre validación: 0.4392968679867519\n",
    "model2.save('baselines/cnn2_64_multi_focal20_100.h5') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b919dbbe0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b905c8e80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'],label=\"Loss\")\n",
    "plt.plot(history.history['val_loss'],label=\"Val Loss\")\n",
    "plt.title(\"Model Loss CNN 2\") \n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.legend(loc=\"upper right\")\n",
    "plt.show()\n",
    "\n",
    "plt.plot(history.history['acc'],label=\"Accuracy\" )\n",
    "plt.plot(history.history['val_acc'],label=\"Val Accuracy\" )\n",
    "plt.title(\"Model Acc CNN 2\") \n",
    "plt.ylabel('acc')\n",
    "plt.xlabel('Epochs')\n",
    "plt.legend(loc=\"upper right\")\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Utilizando recurrentes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "32\n"
     ]
    }
   ],
   "source": [
    "maximo_s=max(map(len,Vec_train)) #91 // probar con 60 (bigote mayor)\n",
    "print (maximo_s)\n",
    "\n",
    "X_train_rnn = sequence.pad_sequences(Vec_train,maxlen=maximo_s,padding='pre',truncating='post',value=np.zeros(100),dtype='float32') \n",
    "X_val_rnn_ig = sequence.pad_sequences(Vec_val_ig,maxlen=maximo_s,padding='pre',truncating='post',value=np.zeros(100),dtype='float32') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def create_simple_GRU(unidades,opt,input_s):\n",
    "    model = Sequential()\n",
    "    model.add(GRU(units=unidades,return_sequences=False,input_shape=input_s))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Dropout(0.3))\n",
    "    model.add(Dense(3, activation='softmax')) \n",
    "    model.compile(loss=focal_loss(2,weights=class_weights_f2), optimizer=opt, metrics=['accuracy'])\n",
    "    return model\n",
    "\n",
    "def create_complex_GRU(unidades,unidades2,opt,input_s):\n",
    "    model = Sequential()\n",
    "    model.add(GRU(units=unidades,return_sequences=True,input_shape=input_s))\n",
    "    model.add(Dropout(0.3))\n",
    "    #model.add(BatchNormalization())\n",
    "    model.add(GRU(units=unidades2,return_sequences=False,input_shape=input_s))\n",
    "    model.add(Dropout(0.2))\n",
    "    model.add(Dense(3, activation='softmax')) \n",
    "    model.compile(loss=focal_loss(2,weights=class_weights_f2), optimizer=opt, metrics=['accuracy'])\n",
    "    #history = model.fit(X_train, y_train, epochs=10, batch_size=128)\n",
    "    return model\n",
    "\n",
    "def create_complex_GRU_2(unidades,unidades2,opt,input_s):\n",
    "    #sgd=optimizers.SGD(lr=0.001, decay=1e-6, momentum=0.9)\n",
    "    model = Sequential()\n",
    "    model.add(GRU(units=unidades,return_sequences=True,input_shape=input_s))\n",
    "    model.add(Dropout(0.45))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(GRU(units=unidades2,return_sequences=False,input_shape=input_s ))#, recurrent_regularizer=regularizers.l1(0.01), kernel_regularizer=regularizers.l2(0.01) ))\n",
    "    model.add(Dropout(0.2))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Dense(3, activation='softmax')) # kernel_regularizer=regularizers.l2(0.01),activity_regularizer=regularizers.l1(0.01))\n",
    "    model.compile(loss=focal_loss(2,weights=class_weights_f2), optimizer=opt, metrics=['accuracy'])\n",
    "    #history = model.fit(X_train, y_train, epochs=10, batch_size=128)\n",
    "    return model\n",
    "\n",
    "def create_complex_GRU_3(unidades,unidades2,unidades3,opt,input_s):\n",
    "    #256,128,64\n",
    "    model3 = Sequential()\n",
    "    model3.add(GRU(units=unidades,return_sequences=True,input_shape=input_s))\n",
    "    model3.add(Dropout(0.65))\n",
    "    model3.add(BatchNormalization())\n",
    "    model3.add(GRU(units=unidades2,return_sequences=True,input_shape=input_s ))#, recurrent_regularizer=regularizers.l1(0.01), kernel_regularizer=regularizers.l2(0.01) ))\n",
    "    model3.add(Dropout(0.45))\n",
    "    model3.add(BatchNormalization())\n",
    "    model3.add(GRU(units=unidades3,return_sequences=False,input_shape=input_s ))#,\n",
    "    model3.add(Dropout(0.35))\n",
    "    model3.add(BatchNormalization())\n",
    "    model3.add(Dense(3, activation='softmax')) # kernel_regularizer=regularizers.l2(0.01),activity_regularizer=regularizers.l1(0.01))\n",
    "    model3.compile(loss=focal_loss(2,weights=class_weights_f2), optimizer=opt, metrics=['accuracy'])\n",
    "    #model3.fit(X_train,y_train,epochs=20,batch_size=64,validation_split=0.25,class_weight=class_weights)\n",
    "    return model3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy sobre validacion: 0.7563291139240507\n",
      "F1-score None sobre validación: [0.29850746 0.2        0.86346154]\n",
      "F1-score macro sobre validación: 0.45398966704936855\n",
      "Precision None sobre validación: [0.31746032 0.16666667 0.87184466]\n",
      "Recall None sobre validación: [0.28169014 0.25       0.8552381 ]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b8720bf28>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "models1 = create_simple_GRU(64,\"adam\",X_train_rnn[0].shape) #modificar post experimentación\n",
    "historys1 = models1.fit(X_train_rnn, y_train_mul,epochs=20,batch_size=64,verbose=0, validation_data=(X_val_rnn_ig, y_val_mul))   \n",
    "\n",
    "trainPredict = models1.predict_classes(X_val_rnn_ig, batch_size=64)\n",
    "acc_val= accuracy_score(y_val_mu, trainPredict)\n",
    "f1_val=f1_score(y_val_mu, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_mu, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_mu, trainPredict, average=None) \n",
    "pre=precision_score(y_val_mu, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_mu, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)\n",
    "\n",
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),['Indirect','Physical','Sexual'],\"Confusion Matrix Caso Multiclass + Focal Loss RNN Simple 100dim\")\n",
    "models1.save('baselines/rnn_simple_multi_focal20_100.h5') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy sobre validacion: 0.8401898734177216\n",
      "F1-score None sobre validación: [0.23913043 0.19047619 0.91327434]\n",
      "F1-score macro sobre validación: 0.4476269871806617\n",
      "Precision None sobre validación: [0.52380952 0.66666667 0.85289256]\n",
      "Recall None sobre validación: [0.15492958 0.11111111 0.98285714]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b8c35d748>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "modelr1 = create_complex_GRU(128,64,\"adam\",X_train_rnn[0].shape) #modificar post experimentación\n",
    "history1 = modelr1.fit(X_train_rnn, y_train_mul,epochs=20,batch_size=64,verbose=0, validation_data=(X_val_rnn_ig, y_val_mul))   \n",
    "\n",
    "trainPredict = modelr1.predict_classes(X_val_rnn_ig, batch_size=64)\n",
    "acc_val= accuracy_score(y_val_mu, trainPredict)\n",
    "f1_val=f1_score(y_val_mu, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_mu, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_mu, trainPredict, average=None) \n",
    "pre=precision_score(y_val_mu, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_mu, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)\n",
    "\n",
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),['Indirect','Physical','Sexual'],\"Confusion Matrix Caso Multiclass + Focal Loss RNN 1 100dim\")\n",
    "modelr1.save('baselines/rnn1_multi_focal20_100.h5') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2 2 2 0 1]\n",
      "Accuracy sobre validacion: 0.7674050632911392\n",
      "F1-score None sobre validación: [0.45238095 0.16091954 0.87215064]\n",
      "F1-score macro sobre validación: 0.49515037893767255\n",
      "Precision None sobre validación: [0.39175258 0.1372549  0.90909091]\n",
      "Recall None sobre validación: [0.53521127 0.19444444 0.83809524]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b85c3c080>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model2 = create_complex_GRU_2(128,64,\"adam\",X_train_rnn[0].shape) #modificar post experimentación\n",
    "history2 = model2.fit(X_train_rnn, y_train_mul,epochs=20,batch_size=64,verbose=0, validation_data=(X_val_rnn_ig, y_val_mul))   \n",
    "\n",
    "trainPredict = model2.predict_classes(X_val_rnn_ig, batch_size=64)\n",
    "print (trainPredict[:5])\n",
    "acc_val= accuracy_score(y_val_mu, trainPredict)\n",
    "f1_val=f1_score(y_val_mu, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_mu, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_mu, trainPredict, average=None) \n",
    "pre=precision_score(y_val_mu, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_mu, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)\n",
    "\n",
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),['Indirect','Physical','Sexual'],\"Confusion Matrix Caso Multiclass + Focal Loss RNN 2 100dim\")\n",
    "model2.save('baselines/rnn2_multi_focal20_100.h5') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b8719c2e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b85738f28>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history2.history['loss'],label=\"Loss\")\n",
    "plt.plot(history2.history['val_loss'],label=\"Val Loss\")\n",
    "plt.title(\"Model Loss RNN 2 100dim\") \n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.legend(loc=\"upper right\")\n",
    "plt.show()\n",
    "\n",
    "plt.plot(history.history['acc'],label=\"Accuracy\" )\n",
    "plt.plot(history.history['val_acc'],label=\"Val Accuracy\" )\n",
    "plt.title(\"Model Acc RNN 2 100dim\") \n",
    "plt.ylabel('acc')\n",
    "plt.xlabel('Epochs')\n",
    "plt.legend(loc=\"upper right\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy sobre validacion: 0.6740506329113924\n",
      "F1-score None sobre validación: [0.43703704 0.14814815 0.79079956]\n",
      "F1-score macro sobre validación: 0.4586615823563615\n",
      "Precision None sobre validación: [0.29648241 0.13333333 0.93041237]\n",
      "Recall None sobre validación: [0.83098592 0.16666667 0.68761905]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f7b80b13898>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model3 = create_complex_GRU_3(256,128,64,\"adam\",X_train_rnn[0].shape) #modificar post experimentación\n",
    "history3 = model3.fit(X_train_rnn, y_train_mul,epochs=20,batch_size=64,verbose=0, validation_data=(X_val_rnn_ig, y_val_mul))   \n",
    "\n",
    "trainPredict = model3.predict_classes(X_val_rnn_ig, batch_size=64)\n",
    "acc_val= accuracy_score(y_val_mu, trainPredict)\n",
    "f1_val=f1_score(y_val_mu, trainPredict, average=None)  # labels=np.unique(trainPredict)\n",
    "f1_val_ma=f1_score(y_val_mu, trainPredict, average='macro')  # labels=np.unique(trainPredict)\n",
    "rec=recall_score(y_val_mu, trainPredict, average=None) \n",
    "pre=precision_score(y_val_mu, trainPredict, average=None) \n",
    "matriz_val=confusion_matrix(y_val_mu, trainPredict)\n",
    "\n",
    "print (\"Accuracy sobre validacion:\",acc_val)  \n",
    "print (\"F1-score None sobre validación:\",f1_val)\n",
    "print (\"F1-score macro sobre validación:\",f1_val_ma)\n",
    "print (\"Precision None sobre validación:\",pre)\n",
    "print (\"Recall None sobre validación:\",rec)\n",
    "\n",
    "plot_confusion_matrix(normalize(matriz_val,axis=1,norm='l1'),['Indirect','Physical','Sexual'],\"Confusion Matrix Caso Multiclass + Focal Loss RNN 2\")\n",
    "model3.save('baselines/rnn3_multi_focal20_100.h5') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.save('matrices/X_train_max.npy',X_train) #fase 1\n",
    "np.save('matrices/X_train_f2_max.npy',X_train_f2)  #fase 2\n",
    "np.save('matrices/X_train_rnn_max.npy',X_train_rnn)  #fase 2\n",
    "np.save('matrices/X_val_ig_max.npy',X_val_ig)\n",
    "np.save('matrices/X_val_rnn_ig.npy',X_val_rnn_ig)\n",
    "\n",
    "np.save('matrices/y_train_bi_max.npy',y_train_bi)\n",
    "np.save('matrices/y_train_bic_max.npy', y_train_bic)\n",
    "np.save('matrices/y_train_mul_max.npy',y_train_mul)\n",
    "\n",
    "np.save('matrices/y_val_bic_max.npy',y_val_bic)\n",
    "np.save('matrices/y_val_bin_max.npy',y_val_bin)\n",
    "np.save('matrices/y_val_mu_max.npy',y_val_mu)\n",
    "np.save('matrices/y_val_mul_max.npy',y_val_mul)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Carga de testing set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2123\n"
     ]
    }
   ],
   "source": [
    "data_test=[]\n",
    "for tw in df_test['tweet_content']:\n",
    "    data_test.append(cleaner(tw))\n",
    "\n",
    "Vec_test=[]\n",
    "tweets_test=data_test\n",
    "vocab=list(embeddings_index.keys())\n",
    "for post in tweets_test:\n",
    "    temp=[]\n",
    "    for term in post:  \n",
    "        if term in vocab_train:\n",
    "            vector=embeddings_index.get(term)\n",
    "            if vector is None:       \n",
    "                temp.append(np.zeros(100))\n",
    "            else:\n",
    "                temp.append(np.asarray(vector))\n",
    "    Vec_test.append(temp)\n",
    "    \n",
    "list_posts_test=[cleaner(text) for text in df_test[\"tweet_content\"].values]  #train\n",
    "\n",
    "all_test=[]\n",
    "for text in list_posts_test:\n",
    "    temp=[]\n",
    "    for word in text:\n",
    "        try:\n",
    "            temp.append(word_index_f2[word])\n",
    "        except:\n",
    "            continue\n",
    "    all_test.append(temp)\n",
    "    \n",
    "x_new_test=all_test#.copy()\n",
    "X_test = sequence.pad_sequences(x_new_test,maxlen=max_input_lenght,padding='post',value=0)\n",
    "    \n",
    "Vec_test_rnn=[]\n",
    "vocab=list(embeddings_index.keys())\n",
    "for post in list_posts_val:\n",
    "    temp=[]\n",
    "    for term in post:   #para todos los terminos de un tweet \n",
    "        if term in vocab_train_f2:\n",
    "            vector=embeddings_index.get(term)\n",
    "            if vector is None:       \n",
    "                temp.append(np.zeros(100))\n",
    "            else:\n",
    "                temp.append(np.asarray(vector))\n",
    "    Vec_test_rnn.append(temp)\n",
    "    \n",
    "all_test_rnn=[]\n",
    "for text in list_posts_test:\n",
    "    temp=[]\n",
    "    for word in text:\n",
    "        try:\n",
    "            temp.append(word_index_f2[word])\n",
    "        except:\n",
    "            continue\n",
    "    all_test_rnn.append(temp)\n",
    "    \n",
    "x_new_test=all_test_rnn#.copy()\n",
    "X_test_rnn = sequence.pad_sequences(x_new_test,maxlen=max_input_lenght,padding='post',value=0)\n",
    "X_test_rnn = sequence.pad_sequences(Vec_test_rnn,maxlen=maximo_s,padding='pre',truncating='post',value=np.zeros(100),dtype='float32') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 2, 0, ..., 2, 2, 2])"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.predict(X_test, batch_size=64).argmax(axis=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.save('matrices/X_test_rnn_max.npy',X_test_rnn)\n",
    "np.save('matrices/X_test_max.npy',X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 230,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2123\n"
     ]
    }
   ],
   "source": [
    "#clases_predichas=model3.predict_classes(X_test_rnn, batch_size=64)\n",
    "#sns.countplot(clases_predichas)"
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}