train_simple_file_enc.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install tensorflow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers, models, callbacks\n",
    "\n",
    "\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "import pandas as pd\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.feature_extraction.text import TfidfTransformer\n",
    "from sklearn.metrics import mean_absolute_error\n",
    "\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "def string_number(x):\n",
    "    x = x.lower()\n",
    "    new = \"\"\n",
    "    for each in x :\n",
    "        new = new + str( ord(each) )\n",
    "    return new\n",
    "\n",
    "\n",
    "def char_number(x):    \n",
    "    return str( ord(x)  )\n",
    "\n",
    "\n",
    "\n",
    "def getConvertedFormatofText(x):\n",
    "\n",
    "    count_vect = CountVectorizer()\n",
    "    X_train_counts = count_vect.fit_transform(x.values.astype('U'))\n",
    "    tfidf_transformer = TfidfTransformer()\n",
    "    X_train_tfidf = tfidf_transformer.fit_transform(X_train_counts)\n",
    "\n",
    "    return tfidf_transformer\n",
    "\n",
    "\n",
    "def getValidationData(xtest, ytest):\n",
    "    xtest, xvalid, ytest, yvalid = train_test_split(xtest, ytest, test_size=0.50, random_state=42)\n",
    "    return xtest, xvalid, ytest, yvalid\n",
    "\n",
    "\n",
    "def getWordSplitData():\n",
    "    data = pd.read_csv(\"word_split_encoded.tsv\", sep=\"\\t\")\n",
    "    data = data.iloc[0:1000000,]\n",
    "    data.columns = [\"words\", \"output\", \"wordlen\"]\n",
    "    #     data = data.dropna()    \n",
    "    x = data.words.map(char_number)\n",
    "    x = x.values.astype('int32')\n",
    "    y = data.output\n",
    "    y = y.values.astype('int64')\n",
    "    return y, x\n",
    "\n",
    "\n",
    "def getWordSplitOHEData():\n",
    "    data = pd.read_csv(\"word_splits_ohe.tsv\", sep=\"\\t\")\n",
    "#     data = data.dropna()\n",
    "    l = [x for x in range(2, data.shape[1])]\n",
    "    x = data.iloc[:,l]\n",
    "    x = x.values.astype('int16')\n",
    "    y = data.output\n",
    "    y = y.values.astype('int64')\n",
    "    return x, y\n",
    "\n",
    "\n",
    "def getTrainTestData():        \n",
    "    x, y = getWordSplitData()\n",
    "    xtrain, xtest, ytrain, ytest = train_test_split(x, y, test_size = 0.40, random_state = 42)\n",
    "    xtest, xvalid, ytest, yvalid = getValidationData(xtest, ytest)\n",
    "    return  xtrain, xtest, xvalid, ytrain, ytest, yvalid\n",
    "\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "def getTheNetwork(num=6):\n",
    "    i = 0\n",
    "    list_dense = []    \n",
    "    list_dense.append(layers.Dense(units= 100 * (num) , activation='relu', input_dim=1))\n",
    "#     list_dense.append(layers.Dropout(0.3))\n",
    "    for x in range(2,num):\n",
    "        i = i+1\n",
    "        list_dense.append(layers.Dense(units= 100 * ((num)-i), activation='relu' ))        \n",
    "#         list_dense.append(layers.Dropout(0.3))\n",
    "        \n",
    "    \n",
    "    list_dense.append(layers.Dense(units= 100, activation='relu' ))\n",
    "    list_dense.append(layers.Dense(units=1))\n",
    "\n",
    "    return  list_dense\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.Sequential(getTheNetwork(6))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_7 (Dense)              (None, 600)               1200      \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 500)               300500    \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 400)               200400    \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 300)               120300    \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 200)               60200     \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 100)               20100     \n",
      "_________________________________________________________________\n",
      "dense_13 (Dense)             (None, 1)                 101       \n",
      "=================================================================\n",
      "Total params: 702,801\n",
      "Trainable params: 702,801\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "\n",
    "\n",
    "model.compile(optimizer='adam', loss = 'huber', metrics=['mse', 'mae', 'mape'])\n",
    "\n",
    "\n",
    "early_stopping = EarlyStopping(\n",
    "    min_delta=0.01,\n",
    "    patience=40,\n",
    "    restore_best_weights=True\n",
    ")\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "xtrain, xtest, xvalid, ytrain, ytest, yvalid = getTrainTestData()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([1011100000000001, 1010105000000001, 1000000011700001,\n",
       "        1000001010800001, 1000011500000001]),\n",
       " array([102, 110, 106, 114, 104], dtype=int32))"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xtrain[0:5], ytrain[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "export_path_keras = \"simple_file_hacker_100.h5\"\n",
    "save_checkpoint = keras.callbacks.ModelCheckpoint(\n",
    "    export_path_keras, monitor='loss', verbose=1, save_best_only=True,\n",
    "    save_weights_only=False, mode='min', save_freq='epoch'\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#char counts train data\n",
    "\n",
    "history = model.fit(\n",
    "    xtrain, ytrain,\n",
    "    validation_data=(xvalid, yvalid),\n",
    "    batch_size=100,\n",
    "    epochs=100,\n",
    "    callbacks=[early_stopping, save_checkpoint],\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "# export_path_keras = \"simple_file_hacker.h5\"\n",
    "# model.save(export_path_keras)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#ohe data trained\n",
    "\n",
    "history = model.fit(\n",
    "    xtrain, ytrain,\n",
    "    validation_data=(xvalid, yvalid),\n",
    "    batch_size=50,\n",
    "    epochs=10,\n",
    "    callbacks=[early_stopping],\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "export_path_keras = \"simple_file_hacker_ohe.h5\"\n",
    "model.save(export_path_keras)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 1000)              2000      \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 900)               900900    \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 800)               720800    \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 700)               560700    \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 600)               420600    \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 500)               300500    \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 400)               200400    \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 300)               120300    \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 200)               60200     \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 100)               20100     \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 1)                 101       \n",
      "=================================================================\n",
      "Total params: 3,306,601\n",
      "Trainable params: 3,306,601\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = models.load_model('simple_file_hacker.h5')\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#char counts\n",
    "history = model.fit(\n",
    "    xtrain, ytrain,\n",
    "    validation_data=(xvalid, yvalid),\n",
    "    batch_size=100,\n",
    "    epochs=50,\n",
    "    callbacks=[early_stopping],\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "export_path_keras = \"simple_file_hacker.h5\"\n",
    "model.save(export_path_keras)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "8.57765388379088"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ypred = model.predict(xtest)\n",
    "\n",
    "mean_absolute_error(ytest, ypred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([1101070000000001, 1000000000011701, 1001010500000001,\n",
       "        1110000000020001, 1000001000170001]),\n",
       " array([[106.939255],\n",
       "        [106.939255],\n",
       "        [106.939255],\n",
       "        [106.939255],\n",
       "        [106.939255]], dtype=float32),\n",
       " array([109, 108, 111,  55, 101], dtype=int32))"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xtest[0:5], ypred[0:5], ytest[0:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.load_weights('simple_file_hacker_ohe.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.load_weights('simple_file_hacker_ohe_v2_01.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 100)               3700      \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 300)               30300     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 400)               120400    \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 500)               200500    \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 600)               300600    \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 700)               420700    \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 1)                 701       \n",
      "=================================================================\n",
      "Total params: 1,076,901\n",
      "Trainable params: 1,076,901\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#char counts\n",
    "history = model.fit(\n",
    "    xtrain, ytrain,\n",
    "    validation_data=(xvalid, yvalid),\n",
    "    batch_size=100,\n",
    "    epochs=10,\n",
    "    callbacks=[early_stopping],\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "export_path_keras = \"simple_file_hacker.h5\"\n",
    "model.save(export_path_keras)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#ohe\n",
    "history = model.fit(\n",
    "    xtrain, ytrain,\n",
    "    validation_data=(xvalid, yvalid),\n",
    "    batch_size=50,\n",
    "    epochs=10,\n",
    "    callbacks=[early_stopping],\n",
    "    verbose=1\n",
    ")\n",
    "\n",
    "export_path_keras = \"simple_file_hacker_ohe_v2_01.h5\"\n",
    "model.save(export_path_keras)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "ypred = model.predict(xtest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[106.773796]], dtype=float32)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ypred = model.predict([[1000000011200001]])\n",
    "\n",
    "ypred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "ypred = model.predict([[1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]])\n",
    "\n",
    "# pd.get_dummies(\"a\", prefix='char')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    [[1000466000000000.0]]\n",
       "dtype: object"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# ypred.values.astype('int64')\n",
    "ypred = pd.Series([ypred])\n",
    "ypred.values.astype('int64')\n",
    "\n",
    "ypred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1064260, 1)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ypred.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ypred = ypred.reshape(1064260, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import classification_report,confusion_matrix\n",
    "\n",
    "print(classification_report(ytest,ypred))\n",
    "print(confusion_matrix(ytest,ypred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluate the model\n",
    "loss, acc = model.evaluate(xtest, ytest, verbose=2)\n",
    "print(\"model, accuracy: {:5.2f}%\".format(100 * acc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "8.548263428280345"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_absolute_error(ytest, ypred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "plt.plot(xtrain,ytrain, color ='r',label = 'Trained data')\n",
    "\n",
    "plt.ylabel('Y axis')\n",
    "plt.xlabel('X axis')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()\n",
    "\n",
    "\n",
    "plt.plot(xtest,ytest, color ='r',label = 'Test data')\n",
    "\n",
    "\n",
    "plt.ylabel('Y axis')\n",
    "plt.xlabel('X axis')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()\n",
    "\n",
    "\n",
    "\n",
    "plt.plot(xtest,ypred, color ='g',label = 'Predicted data')\n",
    "\n",
    "\n",
    "plt.ylabel('Y axis')\n",
    "plt.xlabel('X axis')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8.557379722595215\n",
      "8.557379722595215\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "history_df = pd.DataFrame(history.history)\n",
    "history_df.loc[5:, ['loss', 'val_loss']].plot()\n",
    "history_df.loc[5:, ['mae', 'val_mae']].plot()\n",
    "\n",
    "print(history_df['val_loss'].min())\n",
    "print(history_df['val_mae'].min())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#convert a word to single char cell\n",
    "\n",
    "def writeRow(char, enc, wlen):\n",
    "     with open(\"word_split_encoded.tsv\", 'a') as decodeMsgWritter:\n",
    "        decodeMsgWritter.write(\"\\n{0}\\t{1}\\t{2}\".format(char,enc,wlen))\n",
    "\n",
    "\n",
    "def convertDatato_chars_encryption(x):\n",
    "    row_x = x.words; row_y = x.output;\n",
    "    row_x = list(row_x.lower())\n",
    "    for each in range(16,len(row_y)+16,16):\n",
    "        if each-16 == 0:\n",
    "            writeRow(row_x[int(each/16)-1],row_y[each-16:each], len(row_x) )            \n",
    "        else:\n",
    "            writeRow(row_x[int(each/16)-1],row_y[each-16:each], 0 )\n",
    "\n",
    "    \n",
    "    \n",
    "    \n",
    "data = pd.read_csv(\"title_simple_encoded.tsv\", sep=\"\\t\")\n",
    "data.columns=[\"words\",\"output\"]\n",
    "data = data.dropna()\n",
    "\n",
    "\n",
    "data.apply(convertDatato_chars_encryption, axis = 1)\n",
    "print('completed')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>output</th>\n",
       "      <th>wordlen</th>\n",
       "      <th>char_0</th>\n",
       "      <th>char_1</th>\n",
       "      <th>char_2</th>\n",
       "      <th>char_3</th>\n",
       "      <th>char_4</th>\n",
       "      <th>char_5</th>\n",
       "      <th>char_6</th>\n",
       "      <th>char_7</th>\n",
       "      <th>...</th>\n",
       "      <th>char_q</th>\n",
       "      <th>char_r</th>\n",
       "      <th>char_s</th>\n",
       "      <th>char_t</th>\n",
       "      <th>char_u</th>\n",
       "      <th>char_v</th>\n",
       "      <th>char_w</th>\n",
       "      <th>char_x</th>\n",
       "      <th>char_y</th>\n",
       "      <th>char_z</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1000000101050001</td>\n",
       "      <td>6</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1 rows × 38 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             output  wordlen  char_0  char_1  char_2  char_3  char_4  char_5  \\\n",
       "0  1000000101050001        6       0       0       0       0       0       0   \n",
       "\n",
       "   char_6  char_7  ...  char_q  char_r  char_s  char_t  char_u  char_v  \\\n",
       "0       0       0  ...       0       0       1       0       0       0   \n",
       "\n",
       "   char_w  char_x  char_y  char_z  \n",
       "0       0       0       0       0  \n",
       "\n",
       "[1 rows x 38 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "import pandas as pd\n",
    "\n",
    "data = pd.read_csv(\"word_splits_ohe.tsv\", sep=\"\\t\")\n",
    "data.head(1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#one hot encoder \n",
    "\n",
    "\n",
    "from sklearn.preprocessing import OneHotEncoder as ohe\n",
    "\n",
    "\n",
    "ohei = ohe(handle_unknown='ignore', sparse=False)\n",
    "\n",
    "reshaped_char_data = data.loc[:,[\"char\"]].values.reshape(-1, 1)\n",
    "\n",
    "transformData = pd.DataFrame(ohei.fit_transform(reshaped_char_data))\n",
    "\n",
    "\n",
    "# transformData.index = data.index\n",
    "\n",
    "transformData.head(20)\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.get_dummies(df['country'], prefix='country')\n",
    "\n",
    "transformData.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "# pd.get_dummies(data[\"char\"], prefix='char')\n",
    "\n",
    "\n",
    "data = pd.concat([data ,pd.get_dummies(data['char'], prefix='char')],axis=1)\n",
    "data.drop(['char'], axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data.to_csv(\"word_splits_ohe.tsv\", sep=\"\\t\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "data.dtypes\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install sklearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "conda_mxnet_latest_p37",
   "language": "python",
   "name": "conda_mxnet_latest_p37"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}