Model Training.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.utils import to_categorical\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from keras.layers import LSTM\n",
    "from keras.layers import Embedding\n",
    "\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "# Our Helper Functions\n",
    "from helper import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Read Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "lines = pd.read_pickle(\"./data/02_bin/01_lines.bin\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>user</th>\n",
       "      <th>sess_id</th>\n",
       "      <th>line_id</th>\n",
       "      <th>line_text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>USER1+0</td>\n",
       "      <td>USER0-&gt;0</td>\n",
       "      <td>0</td>\n",
       "      <td>whoami</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>USER1+0</td>\n",
       "      <td>USER0-&gt;0</td>\n",
       "      <td>1</td>\n",
       "      <td>pwd</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>USER1+0</td>\n",
       "      <td>USER0-&gt;0</td>\n",
       "      <td>2</td>\n",
       "      <td>ls</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>USER1+0</td>\n",
       "      <td>USER0-&gt;0</td>\n",
       "      <td>3</td>\n",
       "      <td>dir</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>USER1+0</td>\n",
       "      <td>USER0-&gt;0</td>\n",
       "      <td>4</td>\n",
       "      <td>vi</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      user   sess_id  line_id line_text\n",
       "0  USER1+0  USER0->0        0    whoami\n",
       "1  USER1+0  USER0->0        1       pwd\n",
       "2  USER1+0  USER0->0        2        ls\n",
       "3  USER1+0  USER0->0        3       dir\n",
       "4  USER1+0  USER0->0        4        vi"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lines.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tokenize Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "corpus = lines.line_text\n",
    "tokenizer = Tokenizer(lower=False, split=' ', filters=\"\")\n",
    "tokenizer.fit_on_texts(corpus)\n",
    "vocab_size = len(tokenizer.word_index)+1\n",
    "encoded = tokenizer.texts_to_sequences(corpus)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>line_text</th>\n",
       "      <th>encoded</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>whoami</td>\n",
       "      <td>[485]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>pwd</td>\n",
       "      <td>[42]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ls</td>\n",
       "      <td>[3]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>dir</td>\n",
       "      <td>[22]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>vi</td>\n",
       "      <td>[6]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>source &lt;1&gt;</td>\n",
       "      <td>[100, 1]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>source &lt;1&gt;</td>\n",
       "      <td>[100, 1]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>exit</td>\n",
       "      <td>[13]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>whereis &lt;1&gt;</td>\n",
       "      <td>[293, 1]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>mkdir &lt;1&gt;</td>\n",
       "      <td>[74, 1]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     line_text   encoded\n",
       "0       whoami     [485]\n",
       "1          pwd      [42]\n",
       "2           ls       [3]\n",
       "3          dir      [22]\n",
       "4           vi       [6]\n",
       "5   source <1>  [100, 1]\n",
       "6   source <1>  [100, 1]\n",
       "7         exit      [13]\n",
       "8  whereis <1>  [293, 1]\n",
       "9    mkdir <1>   [74, 1]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corpus.to_frame().assign(encoded=encoded).head(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model 1: One-Word-In, One-Word-Out Sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>line_text</th>\n",
       "      <th>model_inputs</th>\n",
       "      <th>model_inputs_as_nums</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>60</th>\n",
       "      <td>inger &lt;1&gt;</td>\n",
       "      <td>[inger &lt;1&gt;]</td>\n",
       "      <td>[[1177, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61</th>\n",
       "      <td>r -l &lt;2&gt;</td>\n",
       "      <td>[r -l, -l &lt;2&gt;]</td>\n",
       "      <td>[[79, 15], [15, 4]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>62</th>\n",
       "      <td>exit</td>\n",
       "      <td>[exit]</td>\n",
       "      <td>[[13, 0]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>63</th>\n",
       "      <td>elm</td>\n",
       "      <td>[elm]</td>\n",
       "      <td>[[7, 0]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>64</th>\n",
       "      <td>elm</td>\n",
       "      <td>[elm]</td>\n",
       "      <td>[[7, 0]]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    line_text    model_inputs model_inputs_as_nums\n",
       "60  inger <1>     [inger <1>]          [[1177, 1]]\n",
       "61   r -l <2>  [r -l, -l <2>]  [[79, 15], [15, 4]]\n",
       "62       exit          [exit]            [[13, 0]]\n",
       "63        elm           [elm]             [[7, 0]]\n",
       "64        elm           [elm]             [[7, 0]]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corpus.iloc[60:65].to_frame().assign(\n",
    "    model_inputs = [ tokenizer.sequences_to_texts(segment_and_pad(t)) for t in encoded[60:65]],\n",
    "    model_inputs_as_nums = [ segment_and_pad(t) for t in encoded[60:65]]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Segmenting and padding...: 100%|██████████| 145687/145687 [03:41<00:00, 658.52it/s] \n"
     ]
    }
   ],
   "source": [
    "sequences = segment_and_pad_all(encoded, size=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(162049, 2)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sequences = np.array(sequences)s"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = sequences[:,0]\n",
    "y = to_categorical(sequences[:,1], num_classes=vocab_size)\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    X, y, \n",
    "    test_size=0.50, \n",
    "    random_state=2019\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
       "Train on 60768 samples, validate on 20256 samples\n",
      "Epoch 1/10\n",
      "60768/60768 [==============================] - 43s 709us/step - loss: 1.2573 - acc: 0.7394 - val_loss: 0.8739 - val_acc: 0.7904\n",
      "Epoch 2/10\n",
      "60768/60768 [==============================] - 40s 663us/step - loss: 0.7753 - acc: 0.8056 - val_loss: 0.7496 - val_acc: 0.8059\n",
      "Epoch 3/10\n",
      "60768/60768 [==============================] - 41s 682us/step - loss: 0.6865 - acc: 0.8145 - val_loss: 0.6966 - val_acc: 0.8098\n",
      "Epoch 4/10\n",
      "60768/60768 [==============================] - 46s 758us/step - loss: 0.6511 - acc: 0.8172 - val_loss: 0.6836 - val_acc: 0.8107\n",
      "Epoch 5/10\n",
      "60768/60768 [==============================] - 45s 739us/step - loss: 0.6307 - acc: 0.8188 - val_loss: 0.6712 - val_acc: 0.8090\n",
      "Epoch 6/10\n",
      "60768/60768 [==============================] - 41s 680us/step - loss: 0.6177 - acc: 0.8194 - val_loss: 0.6670 - val_acc: 0.8110\n",
      "Epoch 7/10\n",
      "60768/60768 [==============================] - 42s 688us/step - loss: 0.6100 - acc: 0.8197 - val_loss: 0.6614 - val_acc: 0.8092\n",
      "Epoch 8/10\n",
      "60768/60768 [==============================] - 42s 692us/step - loss: 0.6031 - acc: 0.8202 - val_loss: 0.6674 - val_acc: 0.8091\n",
      "Epoch 9/10\n",
      "60768/60768 [==============================] - 41s 682us/step - loss: 0.5984 - acc: 0.8214 - val_loss: 0.6604 - val_acc: 0.8116\n",
      "Epoch 10/10\n",
      "60768/60768 [==============================] - 43s 711us/step - loss: 0.5944 - acc: 0.8220 - val_loss: 0.6597 - val_acc: 0.8073\n"
     ]
    }
   ],
   "source": [
    "model1 = Sequential([\n",
    "    Embedding(vocab_size, 100, input_length=1),\n",
    "    LSTM(100, dropout = 0.2),\n",
    "    Dense(vocab_size, activation='softmax')\n",
    "])\n",
    "model1.compile(\n",
    "    loss='categorical_crossentropy', \n",
    "    optimizer='adam', \n",
    "    metrics=['accuracy']\n",
    ")\n",
    "\n",
    "history = model1.fit(X_train, y_train, validation_split=0.25, epochs=10, batch_size=16, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Validation'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Holdout Set Accuracy = 80.94%\n"
     ]
    }
   ],
   "source": [
    "y_pred = model1.predict_classes(X_test)\n",
    "y_truth = y_test.argmax(axis=1)\n",
    "print(\n",
    "    f\"  Holdout Set Accuracy = {round(accuracy_score(y_truth, y_pred)*100,2)}%\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model 2: Padded Lines"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>line_text</th>\n",
       "      <th>model_inputs_as_nums</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>100</th>\n",
       "      <td>finger &lt;1&gt;</td>\n",
       "      <td>[[0, 0, 18, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>101</th>\n",
       "      <td>lname &lt;2&gt;</td>\n",
       "      <td>[[0, 0, 257, 4]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>102</th>\n",
       "      <td>lname &lt;1&gt; | grep &lt;1&gt;</td>\n",
       "      <td>[[257, 1, 21, 9], [1, 21, 9, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>103</th>\n",
       "      <td>finger &lt;1&gt;</td>\n",
       "      <td>[[0, 0, 18, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>104</th>\n",
       "      <td>finger &lt;1&gt;</td>\n",
       "      <td>[[0, 0, 18, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>105</th>\n",
       "      <td>ping &lt;1&gt;</td>\n",
       "      <td>[[0, 0, 93, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>106</th>\n",
       "      <td>finger &lt;1&gt;</td>\n",
       "      <td>[[0, 0, 18, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>107</th>\n",
       "      <td>finger &lt;1&gt;</td>\n",
       "      <td>[[0, 0, 18, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>108</th>\n",
       "      <td>finger &lt;1&gt;</td>\n",
       "      <td>[[0, 0, 18, 1]]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>109</th>\n",
       "      <td>date</td>\n",
       "      <td>[[0, 0, 0, 59]]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                line_text              model_inputs_as_nums\n",
       "100            finger <1>                   [[0, 0, 18, 1]]\n",
       "101             lname <2>                  [[0, 0, 257, 4]]\n",
       "102  lname <1> | grep <1>  [[257, 1, 21, 9], [1, 21, 9, 1]]\n",
       "103            finger <1>                   [[0, 0, 18, 1]]\n",
       "104            finger <1>                   [[0, 0, 18, 1]]\n",
       "105              ping <1>                   [[0, 0, 93, 1]]\n",
       "106            finger <1>                   [[0, 0, 18, 1]]\n",
       "107            finger <1>                   [[0, 0, 18, 1]]\n",
       "108            finger <1>                   [[0, 0, 18, 1]]\n",
       "109                  date                   [[0, 0, 0, 59]]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corpus.iloc[100:110].to_frame().assign(\n",
    "    model_inputs_as_nums = [ segment_and_pad(t, size=4, pad_pre=False) for t in encoded[100:110]]\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
        "Segmenting and padding...: 100%|██████████| 145687/145687 [03:27<00:00, 701.59it/s]\u001b[A"
     ]
    }
   ],
   "source": [
    "sequences = segment_and_pad_all(encoded, size=4, pad_pre=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "sequences = np.array(sequences)\n",
    "X = sequences[:,:-1]\n",
    "y = to_categorical(sequences[:,-1], num_classes=vocab_size)\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    X, y, \n",
    "    test_size=0.50, \n",
    "    random_state=2019\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 55751 samples, validate on 18584 samples\n",
      "Epoch 1/10\n",
      "55751/55751 [==============================] - 18s 321us/step - loss: 2.9174 - acc: 0.4724 - val_loss: 2.6984 - val_acc: 0.4708\n",
      "Epoch 2/10\n",
      "55751/55751 [==============================] - 15s 277us/step - loss: 2.5750 - acc: 0.4724 - val_loss: 2.2934 - val_acc: 0.4650\n",
      "Epoch 3/10\n",
      "55751/55751 [==============================] - 17s 299us/step - loss: 2.0517 - acc: 0.5916 - val_loss: 1.9543 - val_acc: 0.6241\n",
      "Epoch 4/10\n",
      "55751/55751 [==============================] - 17s 310us/step - loss: 1.8483 - acc: 0.6287 - val_loss: 1.8528 - val_acc: 0.6316\n",
      "Epoch 5/10\n",
      "55751/55751 [==============================] - 16s 284us/step - loss: 1.7598 - acc: 0.6362 - val_loss: 1.7904 - val_acc: 0.6376\n",
      "Epoch 6/10\n",
      "55751/55751 [==============================] - 17s 301us/step - loss: 1.7125 - acc: 0.6390 - val_loss: 1.7553 - val_acc: 0.6406\n",
      "Epoch 7/10\n",
      "55751/55751 [==============================] - 15s 263us/step - loss: 1.6843 - acc: 0.6426 - val_loss: 1.7364 - val_acc: 0.6432\n",
      "Epoch 8/10\n",
      "55751/55751 [==============================] - 15s 268us/step - loss: 1.6652 - acc: 0.6449 - val_loss: 1.7223 - val_acc: 0.6439\n",
      "Epoch 9/10\n",
      "55751/55751 [==============================] - 21s 373us/step - loss: 1.6498 - acc: 0.6471 - val_loss: 1.7085 - val_acc: 0.6474\n",
      "Epoch 10/10\n",
      "55751/55751 [==============================] - 17s 309us/step - loss: 1.6373 - acc: 0.6486 - val_loss: 1.7015 - val_acc: 0.6482\n"
     ]
    }
   ],
   "source": [
    "model2 = Sequential([\n",
    "    Embedding(vocab_size, 10, input_length=3),\n",
    "    LSTM(10, dropout = 0.2),\n",
    "    Dense(vocab_size, activation='softmax')\n",
    "])\n",
    "model2.compile(\n",
    "    loss='categorical_crossentropy', \n",
    "    optimizer='adam', \n",
    "    metrics=['accuracy']\n",
    ")\n",
    "\n",
    "history = model2.fit(X_train, y_train, validation_split=0.25, epochs=10, batch_size=16, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Validation'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Validation'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Holdout Set Accuracy = 64.94%\n"
     ]
    }
   ],
   "source": [
    "y_pred = model2.predict_classes(X_test)\n",
    "y_truth = y_test.argmax(axis=1)\n",
    "print(\n",
    "    f\"  Holdout Set Accuracy = {round(accuracy_score(y_truth, y_pred)*100,2)}%\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Utilization of Results\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict_next(cmd: str) -> str:\n",
    "    return tokenizer.sequences_to_texts([\n",
    "        model1.predict_classes([\n",
    "            tokenizer.word_index[cmd]\n",
    "        ])\n",
    "    ])[0]\n",
    "\n",
    "from random import sample\n",
    "def random_cmd_gen(n: int = 3) -> str:\n",
    "    cur_cmd = sample(tokenizer.word_index.keys(), 1)[0]\n",
    "    full_cmd = cur_cmd\n",
    "\n",
    "    for _ in range(n):\n",
    "        cur_cmd = predict_next(cur_cmd)\n",
    "        full_cmd += f' {cur_cmd}'\n",
    "        if cur_cmd == \"\":\n",
    "            break\n",
    "\n",
    "    return full_cmd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "xrd <1> & cd\n",
      "hi \n",
      ",pre <1> & cd\n",
      "dc \n",
      "gorkemlik \n",
      "lpstat <1> & cd\n",
      "<GENSYM:26>/qshow <2> | grep\n",
      "-kl <1> & cd\n",
      "<GENSYM:1>/hosts <1> & cd\n",
      "psd \n",
      "pw <1> & cd\n",
      ".. \n",
      "makeflat <1> & cd\n",
      "-n2 <1> & cd\n",
      "*.m \n",
      "<GENSYM:0>/talk <1> & cd\n",
      "gorkem <1> & cd\n",
      "dte \n",
      "vdiff \n",
      "audiotool & cd <1>\n"
     ]
    }
   ],
   "source": [
    "for _ in range(20):\n",
    "    print(random_cmd_gen(3))\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:acp2]",
   "language": "python",
   "name": "conda-env-acp2-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": false,
   "sideBar": false,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": false,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}