Model_NLP_and_Change.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import ast\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from sklearn.preprocessing import MultiLabelBinarizer\n",
    "\n",
    "from nltk.stem import WordNetLemmatizer\n",
    "from nltk.tokenize import RegexpTokenizer\n",
    "from nltk.corpus import stopwords\n",
    "from sklearn.feature_extraction.text import CountVectorizer"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Elon Musk's Tweets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [],
   "source": [
    "elon = pd.read_csv('twitter/elonmusk_tweets.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {},
   "outputs": [],
   "source": [
    "elon.text = elon.text.str.lstrip('b')\n",
    "elon.created_at = pd.to_datetime(elon.created_at)\n",
    "elon['date'] = elon.created_at.apply(lambda x: x.date())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {},
   "outputs": [],
   "source": [
    "# NLP processing from the nltk package\n",
    "# Instantiate tokenizer to break text into lists of words\n",
    "#tokenizer = RegexpTokenizer(r'\\w+')\n",
    "tokenizer = RegexpTokenizer(r'[a-zA-Z0-9_@-]+')\n",
    "# Instantiate lemmatizer to turn plural words into singular for better comparison\n",
    "lemmatizer = WordNetLemmatizer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "metadata": {},
   "outputs": [],
   "source": [
    "# First let's turn each title into a list of words\n",
    "elon['nlp_text'] = elon.text.apply(lambda x: tokenizer.tokenize(x.lower()))\n",
    "# Then we'll use the lemmatizer to change plurals to singulars to group them\n",
    "elon.nlp_text = elon.nlp_text.apply(lambda x: [lemmatizer.lemmatize(i) for i in x])\n",
    "# And rejoin the lists of words to one string for the count vectorizer\n",
    "elon.nlp_text = elon.nlp_text.apply(lambda x: ' '.join(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "metadata": {},
   "outputs": [],
   "source": [
    "elon.nlp_text = elon.nlp_text + ' '\n",
    "elon = elon.groupby('date')['nlp_text'].sum()\n",
    "elon = pd.DataFrame(elon)\n",
    "elon.columns = ['elon_tweet']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Tesla company tweets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [],
   "source": [
    "tesla = pd.read_csv('twitter/Tesla_tweets.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "metadata": {},
   "outputs": [],
   "source": [
    "tesla.text = tesla.text.str.lstrip('b')\n",
    "tesla.created_at = pd.to_datetime(tesla.created_at)\n",
    "tesla['date'] = tesla.created_at.apply(lambda x: x.date())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {},
   "outputs": [],
   "source": [
    "# First let's turn each title into a list of words\n",
    "tesla['nlp_text'] = tesla.text.apply(lambda x: tokenizer.tokenize(x.lower()))\n",
    "# Then we'll use the lemmatizer to change plurals to singulars to group them\n",
    "tesla.nlp_text = tesla.nlp_text.apply(lambda x: [lemmatizer.lemmatize(i) for i in x])\n",
    "# And rejoin the lists of words to one string for the count vectorizer\n",
    "tesla.nlp_text = tesla.nlp_text.apply(lambda x: ' '.join(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 177,
   "metadata": {},
   "outputs": [],
   "source": [
    "tesla.nlp_text = tesla.nlp_text + ' '\n",
    "tesla = tesla.groupby('date')['nlp_text'].sum()\n",
    "tesla = pd.DataFrame(tesla)\n",
    "tesla.columns = ['tesla_tweet']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Techmeme headlines"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {},
   "outputs": [],
   "source": [
    "techmeme = pd.read_csv('web_scraping/techmeme.csv', index_col=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 179,
   "metadata": {},
   "outputs": [],
   "source": [
    "techmeme['date_time'] = pd.to_datetime(techmeme.date)\n",
    "techmeme.date = techmeme.date_time.apply(lambda x: x.date())\n",
    "techmeme.extra_sources = techmeme.extra_sources.apply(lambda x: ast.literal_eval(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 180,
   "metadata": {},
   "outputs": [],
   "source": [
    "techmeme['sources'] = techmeme.extra_sources.copy()\n",
    "# Ben helped me with this lovely for loop...\n",
    "for i, list_ in enumerate(techmeme.sources):\n",
    "    list_.append(techmeme.original_source[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "metadata": {},
   "outputs": [],
   "source": [
    "techmeme.original_title = techmeme.original_title.apply(lambda x: x.replace('\\n', ''))\n",
    "techmeme.extra_titles = techmeme.extra_titles.apply(lambda x: x.lstrip('[').rstrip(']'))\n",
    "techmeme['titles'] = techmeme.original_title + ' ' + techmeme.extra_titles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [],
   "source": [
    "# First let's turn each title into a list of words\n",
    "techmeme['nlp_text'] = techmeme.titles.apply(lambda x: tokenizer.tokenize(x.lower()))\n",
    "# Then we'll use the lemmatizer to change plurals to singulars to group them\n",
    "techmeme.nlp_text = techmeme.nlp_text.apply(lambda x: [lemmatizer.lemmatize(i) for i in x])\n",
    "# And rejoin the lists of words to one string for the count vectorizer\n",
    "techmeme.nlp_text = techmeme.nlp_text.apply(lambda x: ' '.join(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [],
   "source": [
    "techmeme.columns = ['date', 'original_source', 'original_title', 'extra_sources',\n",
    "       'extra_titles', 'date_time', 'news_sources', 'titles', 'news_text']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "metadata": {},
   "outputs": [],
   "source": [
    "techmeme.news_text = techmeme.news_text + ' '\n",
    "techmeme = techmeme.groupby('date')['news_sources', 'news_text'].sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Combine them onto stock data\n",
    "\n",
    "In merging, I noticed that I was not retaining the same amounts of information.  I realized that since I do not have stock information on weekends, I have nothing to link tweets or news that happens over the weekend.  I should push this to help predict Monday prices, but I am not yet sure how to do that."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock = pd.read_csv('TESLAHistoricalQuotes.csv')\n",
    "stock = stock.drop(0,0)\n",
    "stock.date = pd.to_datetime(stock.date)\n",
    "stock.date = stock.date.apply(lambda x: x.date())\n",
    "stock = stock.sort_values('date')\n",
    "stock.volume = stock.volume.astype(float)\n",
    "stock.set_index(stock.date, inplace=True)\n",
    "stock = stock.drop('date', 1)\n",
    "stock = pd.merge(stock, techmeme, how='left', left_index=True, right_index=True)\n",
    "stock = pd.merge(stock, tesla, how='left', left_index=True, right_index=True)\n",
    "stock = pd.merge(stock, elon, how='left', left_index=True, right_index=True)\n",
    "stock['daily_gain'] = stock.close - stock.open\n",
    "stock['daily_change'] = stock.daily_gain / stock.open\n",
    "stock['target'] = stock.daily_gain.shift(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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>close</th>\n",
       "      <th>volume</th>\n",
       "      <th>open</th>\n",
       "      <th>high</th>\n",
       "      <th>low</th>\n",
       "      <th>news_sources</th>\n",
       "      <th>news_text</th>\n",
       "      <th>tesla_tweet</th>\n",
       "      <th>elon_tweet</th>\n",
       "      <th>daily_gain</th>\n",
       "      <th>daily_change</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2017-11-16</th>\n",
       "      <td>312.50</td>\n",
       "      <td>5794409.0</td>\n",
       "      <td>313.99</td>\n",
       "      <td>318.14</td>\n",
       "      <td>311.30</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-1.49</td>\n",
       "      <td>-0.004745</td>\n",
       "      <td>-10.62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-11-17</th>\n",
       "      <td>315.05</td>\n",
       "      <td>13720830.0</td>\n",
       "      <td>325.67</td>\n",
       "      <td>326.67</td>\n",
       "      <td>313.15</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>going into plaid http t co ouu9s7cuwi bamf htt...</td>\n",
       "      <td>tesla semi http t co 7vlz7f46ji 0 to 100 km h ...</td>\n",
       "      <td>-10.62</td>\n",
       "      <td>-0.032610</td>\n",
       "      <td>-5.05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-11-20</th>\n",
       "      <td>308.74</td>\n",
       "      <td>8241821.0</td>\n",
       "      <td>313.79</td>\n",
       "      <td>315.50</td>\n",
       "      <td>304.75</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>@schesnais good point</td>\n",
       "      <td>-5.05</td>\n",
       "      <td>-0.016094</td>\n",
       "      <td>6.95</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-11-21</th>\n",
       "      <td>317.81</td>\n",
       "      <td>7256381.0</td>\n",
       "      <td>310.86</td>\n",
       "      <td>318.23</td>\n",
       "      <td>308.71</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>initial hat offering going great w over 300k i...</td>\n",
       "      <td>6.95</td>\n",
       "      <td>0.022357</td>\n",
       "      <td>-4.17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2017-11-22</th>\n",
       "      <td>312.60</td>\n",
       "      <td>4913283.0</td>\n",
       "      <td>316.77</td>\n",
       "      <td>317.42</td>\n",
       "      <td>311.84</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-4.17</td>\n",
       "      <td>-0.013164</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             close      volume    open    high     low news_sources news_text  \\\n",
       "date                                                                            \n",
       "2017-11-16  312.50   5794409.0  313.99  318.14  311.30          NaN       NaN   \n",
       "2017-11-17  315.05  13720830.0  325.67  326.67  313.15          NaN       NaN   \n",
       "2017-11-20  308.74   8241821.0  313.79  315.50  304.75          NaN       NaN   \n",
       "2017-11-21  317.81   7256381.0  310.86  318.23  308.71          NaN       NaN   \n",
       "2017-11-22  312.60   4913283.0  316.77  317.42  311.84          NaN       NaN   \n",
       "\n",
       "                                                  tesla_tweet  \\\n",
       "date                                                            \n",
       "2017-11-16                                                NaN   \n",
       "2017-11-17  going into plaid http t co ouu9s7cuwi bamf htt...   \n",
       "2017-11-20                                                NaN   \n",
       "2017-11-21                                                NaN   \n",
       "2017-11-22                                                NaN   \n",
       "\n",
       "                                                   elon_tweet  daily_gain  \\\n",
       "date                                                                        \n",
       "2017-11-16                                                NaN       -1.49   \n",
       "2017-11-17  tesla semi http t co 7vlz7f46ji 0 to 100 km h ...      -10.62   \n",
       "2017-11-20                             @schesnais good point        -5.05   \n",
       "2017-11-21  initial hat offering going great w over 300k i...        6.95   \n",
       "2017-11-22                                                NaN       -4.17   \n",
       "\n",
       "            daily_change  target  \n",
       "date                              \n",
       "2017-11-16     -0.004745  -10.62  \n",
       "2017-11-17     -0.032610   -5.05  \n",
       "2017-11-20     -0.016094    6.95  \n",
       "2017-11-21      0.022357   -4.17  \n",
       "2017-11-22     -0.013164     NaN  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stock.tail()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Add Change Data as Target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# I will add both and see which makes the most sense\n",
    "stock['daily_gain'] = stock.close - stock.open\n",
    "stock['daily_change'] = stock.daily_gain / stock.open"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We'll use gain first because that is the most useful for my purposes (saving more than $5 is necessary as it is cost of trade)\n",
    "stock['target'] = stock.daily_gain.shift(-1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Use CountVectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "temp_df = stock[stock.news_text.notnull()]\n",
    "cvec = CountVectorizer(stop_words=stopwords.words('english'), ngram_range=(1,2), max_features=1000)\n",
    "cvec.fit(temp_df.news_text.dropna())\n",
    "cvec_df  = pd.DataFrame(cvec.transform(temp_df.news_text.dropna()).todense(), columns=cvec.get_feature_names())\n",
    "cvec_df.index = temp_df.index\n",
    "cvec_df.columns += \" _news_\"\n",
    "stock = stock.join(cvec_df).drop('news_text',1)\n",
    "\n",
    "temp_df = stock[stock.tesla_tweet.notnull()]\n",
    "cvec.fit(temp_df.tesla_tweet.dropna())\n",
    "cvec_df  = pd.DataFrame(cvec.transform(temp_df.tesla_tweet.dropna()).todense(), columns=cvec.get_feature_names())\n",
    "cvec_df.index = temp_df.index\n",
    "cvec_df.columns += \" _tesla_\"\n",
    "stock = stock.join(cvec_df).drop('tesla_tweet',1)\n",
    "\n",
    "temp_df = stock[stock.elon_tweet.notnull()]\n",
    "cvec.fit(temp_df.elon_tweet.dropna())\n",
    "cvec_df  = pd.DataFrame(cvec.transform(temp_df.elon_tweet.dropna()).todense(), columns=cvec.get_feature_names())\n",
    "cvec_df.index = temp_df.index\n",
    "cvec_df.columns += \" _elon_\"\n",
    "stock = stock.join(cvec_df).drop('elon_tweet',1)\n",
    "\n",
    "del temp_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock.news_sources = stock.news_sources.fillna('')\n",
    "mlb = MultiLabelBinarizer()\n",
    "stock = stock.join(pd.DataFrame(mlb.fit_transform(stock.news_sources), columns=mlb.classes_, index=stock.index))\n",
    "stock = stock.drop('news_sources', 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock = stock.fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1866, 3306)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stock.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train/Test Split\n",
    "\n",
    "Now this data is technically not time series anymore, so I ought to be able to do a normal train test split.  It seems like news should have some sort of resiliancy over more days, but I did not see any of that in the EDA."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix, accuracy_score, precision_score, roc_auc_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# I could drop all the prior day info, because I saw no autocorrelation, but it just seems... odd.\n",
    "X_train, X_test, y_train, y_test = train_test_split(stock.drop(['target'], 1), stock['target'], test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LinearRegression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = LinearRegression()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LinearRegression(copy_X=True, fit_intercept=True, n_jobs=1, normalize=False)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import cross_val_score, cross_val_predict\n",
    "from sklearn import metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cross-validated scores: [ -3.84309171e+10  -4.41054402e+10  -4.29125253e+06  -1.14605854e+08\n",
      "  -1.54116437e+10  -1.22452532e+10]\n",
      "Average:  -18385358544.3\n"
     ]
    }
   ],
   "source": [
    "scores = cross_val_score(lr, X_train, y_train, cv=6)\n",
    "print(\"Cross-validated scores:\", scores)\n",
    "print(\"Average: \", scores.mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Score: -4493513377.69\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAakAAAEKCAYAAACopKobAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAIABJREFUeJzt3X+UXHWZ5/H3J50ONo7a/IgMdBKT\nlTZjEAUpY1Z2PYoMCY5rIgsax1kzmjMZHDg6s7sZyDi7OIIjmnHwx4h70DAGV4noQIigtgFk2OMK\npGOQECDSIgPdYSWahEHpCfnx7B/1reR2p7qrurq67+3053VOnb73ud97v09VqurJvfW99yoiMDMz\nK6IpeSdgZmY2FBcpMzMrLBcpMzMrLBcpMzMrLBcpMzMrLBcpMzMrLBcpMzMrLBcpMzMrLBcpMzMr\nrKl5JzDRnXjiiTF79uy80zAzm1A2b978q4iYXqudi9QozZ49m+7u7rzTMDObUCT9Sz3tfLjPzMwK\ny0XKzMwKy0XKzMwKy0XKzMwKy0XKzMwKy6P7zMbZ+i19rO7azo49/ZzS3sbKhXNZcmZH3mmZFVKu\ne1KSnpC0VdIDkrpT7HhJGyU9lv4el+KS9HlJPZIelPT6zHaWpfaPSVqWiZ+Vtt+T1lWjfZg1w/ot\nfay6eSt9e/oJoG9PP6tu3sr6LX15p2ZWSEU43PfWiDgjIkpp/nLgzojoBO5M8wDnA53psQL4EpQL\nDnAF8EZgPnBFpeikNisy6y1qpA+zZlndtZ3+fQcGxPr3HWB11/acMjIrtiIUqcEWA2vT9FpgSSZ+\nQ5TdC7RLOhlYCGyMiF0RsRvYCCxKy14aET+OiABuGLStkfRh1hQ79vSPKG422eVdpAL4gaTNklak\n2EkR8TRA+vvyFO8Ansqs25tiw8V7q8Qb6cOsKU5pbxtR3Gyyy7tInR0Rr6d8mO0SSW8epq2qxKKB\n+HDqWkfSCkndkrp37txZY5Nmh61cOJe21pYBsbbWFlYunJtTRmbFlmuRiogd6e8zwC2Uf1P6ZeUQ\nW/r7TGreC8zMrD4D2FEjPqNKnAb6GJz3dRFRiojS9Ok1r49odsiSMzv45AWn09HehoCO9jY+ecHp\nHt1nNoTchqBLejEwJSKeS9PnAR8HNgDLgKvT31vTKhuASyWtozxI4tmIeFpSF/C3mcES5wGrImKX\npOckLQDuA94PfCGzrbr7GKOXwCapJWd2uCiZ1SnP86ROAm5Jo8KnAt+IiO9L2gTcJGk58CRwUWr/\nXeDtQA/wPPABgFSMrgQ2pXYfj4hdafpDwFeBNuB76QHl4lR3H2Zmlg+VB75Zo0qlUvhWHWZmIyNp\nc+bUoyHlPXDCzMxsSC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5S\nZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWC5SZmZWWLkXKUkt\nkrZIui3Nz5F0n6THJH1T0rQUPybN96TlszPbWJXi2yUtzMQXpViPpMsz8RH3YWZm4y/3IgV8BHgk\nM/8p4JqI6AR2A8tTfDmwOyJOBa5J7ZA0D1gKnAYsAq5Nha8F+CJwPjAPeG9qO+I+zMwsH7kWKUkz\ngD8AvpLmBZwDfDs1WQssSdOL0zxp+dtS+8XAuojYGxG/AHqA+enRExGPR8QLwDpgcYN9mJlZDvLe\nk/os8JfAwTR/ArAnIvan+V6gI013AE8BpOXPpvaH4oPWGSreSB9mZpaD3IqUpHcAz0TE5my4StOo\nsaxZ8Vr9HyJphaRuSd07d+6ssoqZmTVDnntSZwPvlPQE5UNx51Des2qXNDW1mQHsSNO9wEyAtPxl\nwK5sfNA6Q8V/1UAfA0TEdRFRiojS9OnTG3nuZmZWh9yKVESsiogZETGb8sCHuyLifcAPgQtTs2XA\nrWl6Q5onLb8rIiLFl6aReXOATuB+YBPQmUbyTUt9bEjrjLQPMzPLwdTaTcbdZcA6SVcBW4A1Kb4G\n+JqkHsp7N0sBImKbpJuAh4H9wCURcQBA0qVAF9ACXB8R2xrpw8zM8iHvKIxOqVSK7u7uvNMwM5tQ\nJG2OiFKtdnmP7jMzMxuSi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWW\ni5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRWWi5SZmRVWbkVK\n0osk3S/pp5K2SfqbFJ8j6T5Jj0n6pqRpKX5Mmu9Jy2dntrUqxbdLWpiJL0qxHkmXZ+Ij7sPMzMZf\nnntSe4FzIuJ1wBnAIkkLgE8B10REJ7AbWJ7aLwd2R8SpwDWpHZLmAUuB04BFwLWSWiS1AF8Ezgfm\nAe9NbRlpH2Zmlo/cilSU/SbNtqZHAOcA307xtcCSNL04zZOWv02SUnxdROyNiF8APcD89OiJiMcj\n4gVgHbA4rTPSPszMLAe5/iaV9ngeAJ4BNgI/B/ZExP7UpBfoSNMdwFMAafmzwAnZ+KB1hoqf0EAf\nZmaWg1yLVEQciIgzgBmU93xeXa1Z+lttjyaaGB+ujwEkrZDULal7586dVVYxM7NmKMTovojYA9wN\nLADaJU1Ni2YAO9J0LzATIC1/GbArGx+0zlDxXzXQx+B8r4uIUkSUpk+f3tiTNjOzmvIc3TddUnua\nbgPOBR4BfghcmJotA25N0xvSPGn5XRERKb40jcybA3QC9wObgM40km8a5cEVG9I6I+3DzMxyMLV2\nkzFzMrA2jcKbAtwUEbdJehhYJ+kqYAuwJrVfA3xNUg/lvZulABGxTdJNwMPAfuCSiDgAIOlSoAto\nAa6PiG1pW5eNpA8zM8uHvKMwOqVSKbq7u/NOw8xsQpG0OSJKtdoV4jcpMzOzalykzMyssFykzMys\nsFykzMyssFykzMyssFykzMyssFykzMyssFykzMyssFykzMyssFykzMyssFykzMyssFykzMyssFyk\nzMyssFykzMyssFykzMyssOoqUpI+IumlKlsj6SeSzhvr5MzMbHKrd0/qgxHxr8B5wHTgA8DVY5aV\nmZkZ9Rcppb9vB/4xIn6aiTVE0kxJP5T0iKRtkj6S4sdL2ijpsfT3uBSXpM9L6pH0oKTXZ7a1LLV/\nTNKyTPwsSVvTOp+XpEb7MDOz8Vdvkdos6QeUi1SXpJcAB0fZ937gv0XEq4EFwCWS5gGXA3dGRCdw\nZ5oHOB/oTI8VwJegXHCAK4A3AvOBKypFJ7VZkVlvUYqPqA8zM8tHvUVqOeUv8jdExPPANMqH/BoW\nEU9HxE/S9HPAI0AHsBhYm5qtBZak6cXADVF2L9Au6WRgIbAxInZFxG5gI7AoLXtpRPw4IgK4YdC2\nRtKHmZnlYGo9jSLioKRfAvMk1bXOSEiaDZwJ3AecFBFPp36flvTy1KwDeCqzWm+KDRfvrRKngT6e\nHsXTMzOzBtVVcCR9CngP8DBwIIUDuGe0CUj6HeCfgD+PiH9NPxtVbVolFg3Eh02nnnUkraB8OJBZ\ns2bV2KSZmTWq3r2iJcDciNjbzM4ltVIuUF+PiJtT+JeSTk57OCcDz6R4LzAzs/oMYEeKv2VQ/O4U\nn1GlfSN9DBAR1wHXAZRKpVqFz8zMGlTvb1KPA63N7DiNtFsDPBIRf59ZtAGojNBbBtyaib8/jcBb\nADybDtl1AedJOi4NmDgP6ErLnpO0IPX1/kHbGkkfZmaWg3r3pJ4HHpB0J3BobyoiPjyKvs8G/guw\nVdIDKfZXlM+/uknScuBJ4KK07LuURxf2pHw+kHLYJelKYFNq9/GI2JWmPwR8FWgDvpcejLQPMzPL\nh8oD32o0ypx7lBURa6vFJ5NSqRTd3d15p2FmNqFI2hwRpVrt6h3dt1bSNOBVKbQ9IvaNJkEzM7Na\n6h3d9xbK5xM9QXkE3ExJyyJi1KP7zMzMhlLvb1KfAc6LiO0Akl4F3AicNVaJmZmZ1Tu6r7VSoAAi\n4mc0ebSfmZnZYPXuSXVLWgN8Lc2/D9g8NimZmZmV1VukPgRcAnyY8m9S9wDXjlVSZmZmUP/ovr3A\n36eHmZnZuBi2SEm6KSLeLWkrVa5hFxGvHbPMzMxs0qu1J/WR9PcdY52ImZnZYMOO7stct+7PIuJf\nsg/gz8Y+PTMzm8zqHYL++1Vi5zczETMzs8Fq/Sb1Icp7TK+U9GBm0UuA/zuWiZmZmdX6TeoblK8c\n/knKt4+veC5zpXEzM7MxUes3qWcj4gngc8CuzO9R+yS9cTwSNDOzyave36S+BPwmM//bFDMzMxsz\n9RYpRebGUxFxkPqvVmFmZtaQum8fL+nDklrT4yOUbylvZmY2ZuotUhcDbwL6gF7gjcCK0XYu6XpJ\nz0h6KBM7XtJGSY+lv8eluCR9XlKPpAclvT6zzrLU/rHsXYQlnSVpa1rn85LUaB9mZjb+6ipSEfFM\nRCyNiJdHxEkR8YcR8UwT+v8qsGhQ7HLgzojoBO7k8KjC84HO9FhB+k1M0vHAFZQL53zgikrRSW1W\nZNZb1EgfZmaWj1rnSf1lRHxa0heofu2+D4+m84i4R9LsQeHFwFvS9FrgbuCyFL8h/TZ2r6R2SSen\nthsrQ+IlbQQWSbobeGlE/DjFbwCWUB5SP6I+MlfeMDOzcVRr8MMj6W/3WCeScVKlKETE05JenuId\nwFOZdr0pNly8t0q8kT5cpMzMcjBskYqI76S/a8cnnWGpSiwaiDfSx8BG0grSb3KzZs2qsUkzM2tU\nrcN932GYL/aIeGfTM4JfVg6xpcN5ld++eoGZmXYzgB0p/pZB8btTfEaV9o30MUBEXAdcB1AqlWoV\nPjMza1CtgRN/B3wG+AXQD3w5PX4DPDTMeqOxAaiM0FsG3JqJvz+NwFsAPJsO2XUB50k6Lg2YOA/o\nSsuek7Qgjep7/6BtjaQPMzPLQa3Dff8MIOnKiHhzZtF3JN0z2s4l3Uh5L+hESb2UR+ldDdwkaTnw\nJHBRav5d4O1AD/A88IGU4y5JVwKbUruPZ64r+CHKIwjbKA+Y+F6Kj6gPMzPLhzIXkhi6kfQI8AcR\n8XianwN8NyJePcb5FV6pVIru7vEcV2JmNvFJ2hwRpVrt6r200V8Ad0uqXGViNvCnDeZmZmZWl7qK\nVER8X1In8Hsp9GhE7B27tMzMzOosUpKOBf4r8IqI+BNJnZLmRsRtY5uemdnw1m/pY3XXdnbs6eeU\n9jZWLpzLkjM7aq9oE0K91+77R+AF4N+n+V7gqjHJyMysTuu39LHq5q307ekngL49/ay6eSvrt/Tl\nnZo1Sb1F6pUR8WlgH0BE9FP9xFczs3Gzums7/fsODIj17zvA6q7tOWVkzVbvwIkXJLWRTuyV9ErA\nv0mZNYEPVzVux57+EcVt4qm3SF0BfB+YKenrwNnAH49VUmaTReVwVWVvoHK4CnChqsMp7W30VSlI\np7S35ZCNjYWah/vS1RoeBS6gXJhuBEoRcfeYZmY2CRT5cNX6LX2cffVdzLn8ds6++q5C/s6zcuFc\n2lpbBsTaWltYuXBuThlZs9Xck4qIkLQ+Is4Cbh+HnMwmjaIerpooe3iVXHy49OhV7+G+eyW9ISI2\n1W5qZvUq6uGq4fbwilYAlpzZUbicrHnqHd33VsqF6ufptupbJT04lomZjcZEOFQFxT1cVdQ9PJt8\n6t2TOn9MszBroolyqAqKe7iqqHt4NvnUup/Ui4CLgVOBrcCaiNg/HomZNWoiHaqCYh6uWrlw7oBC\nD8XYw7PJp9ae1FrKJ/D+H8p7U/OAj4x1Umaj4UNVo1fUPTybfGoVqXkRcTqApDXA/WOfktno+FBV\ncxRxDw988vNkU2vgxL7KhA/z2URR1MEIR5O8Bqb4Wn2Tz7A3PZR0APhtZZbyHW6fT9MRES8d8wwL\nzjc9LCb/b7v5Kq9p357+8hdAZlllvqOO13rwv81bf286P3x054B/K6h+qPHsq++qupfc0d7Gjy4/\np6nPdyKYyO/zem96WNedeW1oLlITVzM/4GP5ZTGgOAgqH9kpgoNxuDBA7d+QsttqkTgQUXX9l7W1\nIsGe5/cdKiT/tLnviAEp1bS1tvDJC04f0PdwBW6w1iniIHDg4OFWrS1i9YWv4y+++UDVdQVc854z\nJuwXdiMGj2KF6q99tfVG+zo1YxsuUqMgaRHwOaAF+EpEXD1UWxep+mXf2O3HthIBz/bvq/m/53q3\nOZIPy1+v38rX731y2C/LwY6ZOoUX9h889KV9+4NPs/v5fVXbDvVFXCkMg7VOgQMH4eAI8rHhDd67\nA44o0O2pGO9+ft+hWLV/u472Nmaf0Ma9j+8+4t/vxdNa+MS76isM1f5z0GghHWqv8rhjW9nyP887\nou/K5+43/7affZn/AAxV2Ib6bDVaHAdzkWqQpBbgZ8DvU75v1ibgvRHxcLX2k61INVoUqr2xs1pb\nBEFdH57htjl4nfVb+vjYhm3s6S8Xk2NbpyCJ375Qe4/Ajh7V3l/N1DJFfOai11V9r9Z67//Rgllc\nteT0EX221m/p48+/+cCQ+Xz2PWcMWVCqGXy4dLjPVqXY1tpGLfUWqXqvODGZzAd6IuLxiHgBWAcs\nbnYnE+WKCFmj+dG62rlLWfsOxBFfILUutFrr4qzrt/Sx8ls/PVSgAJ7fd9AFahKq9v5qpgMHY8j3\naq33/v++90n+ev3Wuj9blc/hcCq51Oq7YvDpGcN9tsb7FA8XqSN1AE9l5ntT7BBJKyR1S+reuXPn\niDuYqCOURnPF7kbfwMOtV+vDsrpr+5h+MZlljebL+8b7nqr7s1VP4an0We/nbvDpGcM9l6FO5Rir\nUzxcpI5U7Y7DA77pIuK6iChFRGn69Okj7qDIt2cYzmg+hI2+gYdbr9aHxSfv2ngazZd3td8pofp7\neCSft3r6rnZ6xnDPZbxP8XCROlIvMDMzPwPY0cwOJuoVEUbzIaz2xs5qbRGtUwb+/6DWG7/Wh8Un\n71rW4PdXM7VM0ZDv1VrvfSgPqKmm2nu41vu6teVwLtX6bm1RecAI5d+Rqv3uO9xna8mZHXzygtPp\naG8bdhvNUu8FZieTTUCnpDlAH7AU+MNmdjBRr4gwmuu5Db7MTjNG99W6dM/KhXNZ+a2fjskhv8rw\nbys+Ae9bMIvSK47nb76zbcgRmY2qNbqvEh+q79Yp4j3zZx4xxH+oz1a1z2HFcce2csV/Ou1Qn41e\n3qrWeuN5NRKP7qtC0tuBz1Iegn59RHxiqLaNjO5r1hDOPEy0kwerje47prXl0Pk/v927f8DAiopq\no50GP2/giH9HAW965fE88ev+AcV4T//hIc4vntYyosEbAo5N6zS6jaFUtnfcsa38du9+XjhQrO+D\n9rZWPvbO8pfu7MuHvufqHy2YxW0/ffrQv2XlfLJqw7yHOmcr+wX/1+u3cuN9Tx0akn7stBaef+FA\nU86nyxar7PMb6ei+ifQ5rMZD0MdJo0PQj4Y32dFgtP9hGM2Q/MHnzRxXZe9yuG0NdZ7MUOoZIlzt\nahDZL//B/1Ov9vpl91yGOs+mkdfMV5s4urhIjZPJdp7U0Wii/odhqALxplcez0+efHbc9tTH6/Wb\nyEcg7EguUuPERcryNNxVASZi4a3laH1ek5GL1DhxkbIsf4ma1afeIuXRfWZNMpFuW282Ufg8KbMm\nmagnaZsVmYuUWZNM1JO0zYrMRcqsScb7mmZmk4GLlFmT+Lb1Zs3ngRNmTdLoJWjMbGguUmZNNJ7X\nNDObDHy4z8zMCstFyszMCstFyszMCstFyszMCstFyszMCstFyszMCstFyszMCiuXIiXpIknbJB2U\nVBq0bJWkHknbJS3MxBelWI+kyzPxOZLuk/SYpG9Kmpbix6T5nrR8dqN9mJlZPvLak3oIuAC4JxuU\nNA9YCpwGLAKuldQiqQX4InA+MA94b2oL8CngmojoBHYDy1N8ObA7Ik4FrkntGu3DzMxykEuRiohH\nIqLa/QsWA+siYm9E/ALoAeanR09EPB4RLwDrgMWSBJwDfDutvxZYktnW2jT9beBtqf2I+mjuMzcz\ns5Eo2m9SHcBTmfneFBsqfgKwJyL2D4oP2FZa/mxqP9I+jiBphaRuSd07d+4c4VM0M7N6jVmRknSH\npIeqPIbbO1GVWDQQb/a2BgYjrouIUkSUpk+fXq2JmZk1wZhdYDYizm1gtV5gZmZ+BrAjTVeL/wpo\nlzQ17S1l21e21StpKvAyYFcDfZiZWU6KdrhvA7A0jcybA3QC9wObgM40km8a5YEPGyIigB8CF6b1\nlwG3Zra1LE1fCNyV2o+ojzF+vmZmNoxcbtUh6V3AF4DpwO2SHoiIhRGxTdJNwMPAfuCSiDiQ1rkU\n6AJagOsjYlva3GXAOklXAVuANSm+BviapB7Ke1BLARrsw8zMcqDyzoU1qlQqRXd3d95pmJlNKJI2\nR0SpVruiHe4zMzM7xEXKzMwKy0XKzMwKy0XKzMwKy0XKzMwKy0XKzMwKy0XKzMwKy0XKzMwKy0XK\nzMwKK5fLIplZsazf0sfqru3s2NPPKe1trFw4lyVnVr1Tjdm4cpEym+TWb+lj1c1b6d93AIC+Pf2s\nunkrgAuV5c6H+8wmudVd2w8VqIr+fQdY3VXt5tlm48tFymyS27Gnf0Rxs/HkImU2yZ3S3jaiuNl4\ncpEym+RWLpxLW2vLgFhbawsrF87NKSOzwzxwwmySqwyO8Og+KyIXKTNjyZkdLkpWSLkc7pO0WtKj\nkh6UdIuk9syyVZJ6JG2XtDATX5RiPZIuz8TnSLpP0mOSvilpWoofk+Z70vLZjfZhZmb5yOs3qY3A\nayLitcDPgFUAkuYBS4HTgEXAtZJaJLUAXwTOB+YB701tAT4FXBMRncBuYHmKLwd2R8SpwDWpXaN9\nmJlZDnIpUhHxg4jYn2bvBWak6cXAuojYGxG/AHqA+enRExGPR8QLwDpgsSQB5wDfTuuvBZZktrU2\nTX8beFtqP6I+xuL5m5lZfYowuu+DwPfSdAfwVGZZb4oNFT8B2JMpeJX4gG2l5c+m9iPtw8zMcjJm\nAyck3QH8bpVFH42IW1ObjwL7ga9XVqvSPqheTGOY9sNta6R9HEHSCmAFwKxZs6o1MTOzJhizIhUR\n5w63XNIy4B3A2yKiUgx6gZmZZjOAHWm6WvxXQLukqWlvKdu+sq1eSVOBlwG7Guij2nO7DrgOoFQq\nVS1kZmY2enmN7lsEXAa8MyKezyzaACxNI/PmAJ3A/cAmoDON5JtGeeDDhlTcfghcmNZfBtya2day\nNH0hcFdqP6I+xuL5m5lZffI6T+ofgGOAjeWxDNwbERdHxDZJNwEPUz4MeElEHACQdCnQBbQA10fE\ntrSty4B1kq4CtgBrUnwN8DVJPZT3oJYCNNiHmZnlQIePtFkjSqVSdHd3552GmdmEImlzRJRqtSvC\n6D4zM7OqXKTMzKywXKTMzKywXKTMzKywXKTMzKywfKsOs6PU+i19vkeUTXguUmZHofVb+lh181b6\n9x0AoG9PP6tu3grgQmUTig/3mR2FVndtP1SgKvr3HWB11/acMjJrjIuU2VFox57+EcXNispFyuwo\ndEp724jiZkXlImV2FFq5cC5trS0DYm2tLaxcODenjMwa44ETZkehyuAIj+6zic5FyuwoteTMDhcl\nm/B8uM/MzArLRcrMzArLRcrMzArLRcrMzArLRcrMzArLt48fJUk7gX8Z5WZOBH7VhHTGmvNsvomS\nq/NsvomS61jl+YqImF6rkYtUAUjqjohS3nnU4jybb6Lk6jybb6LkmneePtxnZmaF5SJlZmaF5SJV\nDNflnUCdnGfzTZRcnWfzTZRcc83Tv0mZmVlheU/KzMwKy0UqJ5JWS3pU0oOSbpHUnlm2SlKPpO2S\nFuaZZ8rnIknbJB2UVMrEZ0vql/RAevyvIuaZlhXqNa2Q9DFJfZnX8O155zSYpEXpdeuRdHne+QxF\n0hOStqbXsTvvfLIkXS/pGUkPZWLHS9oo6bH097g8c0w5Vcsz1/eoi1R+NgKviYjXAj8DVgFImgcs\nBU4DFgHXSmoZcivj4yHgAuCeKst+HhFnpMfF45zXYFXzLOhrmnVN5jX8bt7JZKXX6YvA+cA84L3p\n9Syqt6bXsWhDu79K+b2XdTlwZ0R0Anem+bx9lSPzhBzfoy5SOYmIH0TE/jR7LzAjTS8G1kXE3oj4\nBdADzM8jx4qIeCQitueZQz2GybNwr+kEMh/oiYjHI+IFYB3l19NGICLuAXYNCi8G1qbptcCScU2q\niiHyzJWLVDF8EPhemu4Ansos602xopojaYukf5b0H/NOZghFf00vTYd9ry/CIZ9Biv7aZQXwA0mb\nJa3IO5k6nBQRTwOkvy/POZ/h5PYe9U0Px5CkO4DfrbLooxFxa2rzUWA/8PXKalXaj/kQzHpyreJp\nYFZE/FrSWcB6SadFxL8WLM9cXtNDnQ+TM/Al4MqUz5XAZyj/p6Uocn3tRujsiNgh6eXARkmPpj0D\nG51c36MuUmMoIs4dbrmkZcA7gLfF4XMBeoGZmWYzgB1jk+FhtXIdYp29wN40vVnSz4FXAWP2o3Uj\neZLTa1pRb86SvgzcNsbpjFSur91IRMSO9PcZSbdQPlRZ5CL1S0knR8TTkk4Gnsk7oWoi4peV6Tze\noz7clxNJi4DLgHdGxPOZRRuApZKOkTQH6ATuzyPHWiRNrwxAkPTvKOf6eL5ZVVXY1zR9OVW8i/Lg\njyLZBHRKmiNpGuUBKBtyzukIkl4s6SWVaeA8ivdaDrYBWJamlwFDHQnIVd7vUe9J5ecfgGMoH5YA\nuDciLo6IbZJuAh6mfBjwkog4kGOeSHoX8AVgOnC7pAciYiHwZuDjkvYDB4CLIyK3H12HyrOIr2nG\npyWdQflQyhPAn+abzkARsV/SpUAX0AJcHxHbck6rmpOAW9JnaSrwjYj4fr4pHSbpRuAtwImSeoEr\ngKuBmyQtB54ELsovw7Ih8nxLnu9RX3HCzMwKy4f7zMyssFykzMyssFykzMyssFykzMyssFykzMys\nsFykzMaBpBMyV5H+f4OuKj2tSX28RNKvJf3OoPhtki4YZr1zJa1vRg5mzebzpMzGQUT8GjgDyrc+\nAH4TEX+XbaPyST6KiIMN9vHY/p+3AAACS0lEQVScpLsoX7j062mbxwFvBC5sPHuz/HhPyixHkk6V\n9JDK9+L6CTBT0p7M8qWSvpKmT5J0s6RuSfdLWlBlkzdSvipExX8Gbo+If5O0QNKP0wWBfySps0o+\nV0n688z8o5JmpOllqd8HJF0raYqkqZK+pvJ9nB6S9OHmvDJmZS5SZvmbB6yJiDOBvmHafR74dLpX\n0ruBr1RpczuwIHOl6qWUCxfAI8B/SP1cCVxVb4KSXkP5kjhviogzKB+FWQqcBZwYEadHxGuAG+rd\nplk9fLjPLH8/j4hNdbQ7F5ibLv0DcJyktojorwQiYq+k24ELJN1G+UaPd6bF7cANkl7ZQI7nAm8A\nulP/bZRv4dGVcvoc8F3gBw1s22xILlJm+fttZvogA2+P8aLMtID56eaDw7kR+O+UC8nNmZtrfgLo\niohrJZ0KVLu23X4GHmGp9C/K1+37H4NXkPRaynfu/TDlw4sT4V5ONkH4cJ9ZgaRBE7sldUqaQvkQ\nW8UdwCWVmXTRz2ruoLwHdTGHD/UBvIzDhxP/eIh1n6B8CA9J8zl8m447gHdLOjEtO0HSLEnTKQ/2\n+Bbli5G+vo6naVY3Fymz4rmM8l7OnZTv51RxCXB2ukPqw8CfVFs5XeH9FuClwI8yiz4FrJb0o2rr\nJd8CTpK0BVhOuvVKRGwF/ga4Q9KDlA/rnUS5iN0j6QHgy8BfjfC5mg3LV0E3M7PC8p6UmZkVlouU\nmZkVlouUmZkVlouUmZkVlouUmZkVlouUmZkVlouUmZkVlouUmZkV1v8HzAonxGJxd1IAAAAASUVO\nRK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x114976cf8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "predictions = lr.predict(X_test)\n",
    "plt.scatter(y_test, predictions)\n",
    "plt.xlabel(\"True Values\")\n",
    "plt.ylabel(\"Predictions\")\n",
    "print(\"Score:\", lr.score(X_test, y_test))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda/lib/python3.6/site-packages/sklearn/cross_validation.py:41: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.\n",
      "  \"This module will be removed in 0.20.\", DeprecationWarning)\n"
     ]
    }
   ],
   "source": [
    "from xgboost import XGBRegressor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "xgb = XGBRegressor(gamma=0, learning_rate=0.02, max_depth=3, n_estimators=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "XGBRegressor(base_score=0.5, colsample_bylevel=1, colsample_bytree=1, gamma=0,\n",
       "       learning_rate=0.02, max_delta_step=0, max_depth=3,\n",
       "       min_child_weight=1, missing=None, n_estimators=100, nthread=-1,\n",
       "       objective='reg:linear', reg_alpha=0, reg_lambda=1,\n",
       "       scale_pos_weight=1, seed=0, silent=True, subsample=1)"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xgb.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Score: -0.0126137236781\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYQAAAEKCAYAAAASByJ7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAPx/MPSwo7qS/cYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11faac2b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "predictions = xgb.predict(X_test)\n",
    "plt.scatter(y_test, predictions)\n",
    "plt.xlabel(\"True Values\")\n",
    "plt.ylabel(\"Predictions\")\n",
    "print(\"Score:\", xgb.score(X_test, y_test))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cross-Predicted Accuracy: -0.0126137236781\n"
     ]
    }
   ],
   "source": [
    "accuracy = metrics.r2_score(y_test, predictions)\n",
    "print(\"Cross-Predicted Accuracy:\", accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(stock.iloc[915:-1].drop(['target'], 1), stock.iloc[915:-1]['target'], test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Score: 0.0170629618254\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYQAAAEKCAYAAAASByJ7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAIABJREFUeJzt3XuYZHV95/H3Z5oGeojag44XBkZG\nJRDuSIuzS9aEhADex/EGq1kTszvLRlfiY1AQN+iDPkwyMW4uugmJRDQE0QijMkbkYsKGBLWHmWFA\nBgUEtGFlFMYbHWxmvvtHnWqqa05VnbqcPudUfV7P0093V50651vVp8/3/O6KCMzMzJYUHYCZmZWD\nE4KZmQFOCGZmlnBCMDMzwAnBzMwSTghmZgYUmBAk7S/p65K2SbpD0geKisXMzEBFjUOQJOCAiPip\npHHgX4BzIuKWQgIyMxtx+xR14Khlop8mv44nXx4lZ2ZWkMISAoCkMWAz8ALgoxHxtZRt1gHrAA44\n4IATjzjiiMUN0sys4jZv3vyDiFjeabvCqowWBCFNAlcD/zMibm+13dTUVExPTy9eYGZmQ0DS5oiY\n6rRdKXoZRcQu4J+AMwoOxcxsZBXZy2h5UjJA0gRwKrCjqHjMzEZdkW0IzwEuS9oRlgCfiYhrCozH\nzGykFdnL6DbghKKOb2ZmC5WiDcHMzIrnhGBmZkDB4xDMhsHGLTNsuPYuHtw1y0GTE5x7+uGsOWFF\n0WGZdc0JwawPG7fMcP5V25md2w3AzK5Zzr9qO4CTglWOq4zM+rDh2rvmk0Hd7NxuNlx7V0ERmfXO\nCcGsDw/umu3qcbMyc0Iw68NBkxNdPW5WZk4IZn049/TDmRgfW/DYxPgY555+eEERmfXOjcpmfag3\nHLuXkQ0DJwSzPq05YYUTgA0FVxmZmRnghGBmZglXGZlZrjySuzqcEMwsNx7JXS2uMjKz3Hgkd7U4\nIZhZbjySu1qcEMwsNx7JXS1OCGaWG4/krhY3KptZbjySu1qcEMwsVx7JXR2uMjIzM6DAhCDpEElf\nlXSnpDsknVNULGZmVmyV0RPAuyLiVklPATZLui4ivllgTGZmI6uwEkJEPBQRtyY//wS4E3BFo5lZ\nQUrRhiDpUOAE4Gspz62TNC1peufOnYsdmpnZyCg8IUj6BeBzwO9FxI+bn4+ISyJiKiKmli9fvvgB\nmpmNiEITgqRxasng8oi4qshYzMxGXZG9jAR8HLgzIv6kqDjMzKymyF5GJwO/CWyXtDV57L0R8aUC\nYzIbGV6nwJoVlhAi4l8AFXV8s1HmdQosTeGNyma2+LxOgaVxQjAbQV6nwNJ4cjuznJS5jv6gyQlm\nUi7+XqdgtLmEYJaDeh39zK5Zgifr6DdumSk6NMDrFFg6JwSzHJS9jn7NCSu4eO0xrJicQMCKyQku\nXntMaUowVgxXGZnloAp19F6nwJq5hGCWA68lbFXkhGCWA9fRWxW5ysgsB15L2KrICcEsJ66jt6px\nlZGZmQFOCGZmlnBCMDMzwAnBzMwSTghmZga4l5EVrMwTwGXRbfxlf79lj8/y5YRghan6Ii3dxl/2\n91v2+Cx/rjKywpR9ArhOuo2/7O+37PFZ/pwQrDBVmACunW7jL/v7LXt8lj8nBCtM1SeA6zb+sr/f\nssdn+Ss0IUi6VNLDkm4vMg4rRtUngOs2/rK/37LHZ/krulH5E8BfAJ8sOA4rQNUngOs2/rK/37LH\nZ/lTRBQbgHQocE1EHN1p26mpqZiens49JjOzYSJpc0RMddrObQhmZgZUICFIWidpWtL0zp07iw7H\nzGxolT4hRMQlETEVEVPLly8vOhwzs6FVdKOymY0IT4tRfkV3O70C+DfgcEnfk/Q7RcZjZvmoT4sx\ns2uWoDYtxjuv3Mr7Nm4vOjRrUGgJISLOKvL4ZrY40qbFCODyWx5g6rkHuqRQEqVvQzCz6ms1/UWA\n50oqEbchmFVQ1erjD5qcYGaE50qqyt/LJQSznGzcMsPJ629k1XmbOHn9jWzcMjOw/TbXx59/1faB\n7T8P555+OGrx3LDPlVSlv5dLCFY6ed5N9bLvXl+T19oC7aapLuNdJ9Te8/T9j3D5LQ/QODfCKMyV\nVKW/l0sIVip53k31su9e48lzbYGqTlP9wTXH8JE3Hs+KyQkErJic4OK1x5TuojhoVfp7uYRgpZLl\nbqrXEkQvd2pZX9McU5715a32X4WqlzUnrBj6BNCsSn8vlxCsVDrdTfVTgujlTi3La9JiyrO+3NNU\nV0uV/l5OCFYqnRZp6acqppcFYFo9t0SaT0Kt+tg3J4VBXQTWnLCCi9ceM3JVL1VVpb+Xq4ysVM49\n/fAFjbGw8ELaT31sp31nfQ3A7oj5RuJ2fexXTE4MrHG8Kl0XbW9VqSpzQrBS6bRISz/1sb0sAFN/\n7l2f2cbuprVD6iWTVjGtmJzg5vN+rWNcWeTZa8msrvAFcrrhBXKs+cIItbv8QRfB63fjM7tmGZP2\nSgZ1Aj7yxuM7xtR8d3/KEcv56o6dmRPTyetvzD3p2PDKukCOSwhWKfWL5ge+eAePPjYHwH77DLYp\nrDnptEoGUCuZdCp5pN3d/90tD8zvI8vdfqtqqZlds6w6b5OrkHIyatV0TghWSf8+t2f+512zc20v\nqN3+U6c1EqdpbH9oV0ecZX+dur+268ra2NuqHov1bxSr6dzLyCqnVU+jD3zxjr2miuilm2qWBupu\neopkHXvQbru0rovNBjXwzWryHFxYVi4hWGF6LY63unA++tjcfDVS/cK/3z5LUv+p3/WZbbzzyq2p\nx213N95o+v5H2lYT1Z9b0qYNolFjw3jaZ3Px2mPmH2u1tyxxt9r/sN719qpKI4wHxSUEy027yd36\nGWCWdXDX7Nxuds3OpT63O6LlcbPcjdfbAdLib35vWZJBY/VTq88G4Obzfo3vrH85Y2o19I2OE+lV\nabK1IvUybqXqnBAsF50uOv0Ux7NcsLvRfNzGgUQAba69qftp1WYwJs0PTHrz6pUtBypl+WzaJZlO\nF/hRrArpRZVGGA+Ku50OSJmK4P3O6Pm0iXEk2PXYXGpVSGMPH4BlS8e58JVHLThGq26SYxIffsNx\nvPPKrS2rPRpNjC9h//ExHn1sDkHL10gwqFN5CTCx7xg/+3nnhuUqGJNY/bxl3PHgT1qWmLoxMb6E\nF66c5JZ7H12QmMYkznrxIQALelHV7Tsm5nbXSmb1baeeeyDv/8IdC+JKO596kdZ1eEWbqr20/5Uy\n/V/3I2u3UyeEAVisvvF5xZL2mkb11wOc+w/bmNu99zkzPiY2vO64+WMcet6mljFOjI+xRAzNBdd6\n1yrJN59P3Wp3Tjeez2X5v81b1oTgKqMBKFMRvJdYOnWLbKwKSUsGAHO7I/P7nZ3bzWNOBkbrEl83\n51Oadud0u6q9Ua86y5QQJJ0j6amq+bikWyWdlndwVVGm3giDnNGzeZtO23XzfqtTLrWi9PP/k+Vc\nLdP/bVlkLSG8NSJ+DJwGLAd+G1jf78ElnSHpLkl3Szqv3/0VpUy9EQY5o2fzNp22WyLN9yha0qEh\ntl0vGTPo7/+n02vbnc/D3Iuok6wJof7f+zLgbyNiG3vP7tsVSWPAR4GXAkcCZ0k6sp99FqVMvRF6\niaVTr5366889/XDGx1r/2Ru7crYrAkyMj3HWiw8ZaE8hq6ZWZ9P4mPr6/2l3Tjeez2X5vy2LrAPT\nNkv6CrAKOF/SU4A9HV7TyUnA3RFxL4CkTwOvBr7Z534XXS+zaJYplubXtOtlBOzVyyith88ear1R\nfv5ELOiJ0tjLY+q5B+7VC2SJYE/DvrL2Mhok9zJqrwq9jBrP6Xa9jOrbFP1/WxaZehlJWgIcD9wb\nEbskPR1YERG39Xxg6XXAGRHxX5PffxN4cUS8vWm7dcA6gJUrV554//3393pIy8mq8zalXqgFfGf9\ny3M7bjczgPYaY3O3w0d/9jiPze19L9Rq1tFWx+0UQ6v3lkXjRa9Vt8mi/mZWjIH2MoqIPcD3gSMl\nvQQ4CpjsL8TU0uJe52hEXBIRUxExtXz58j4PaXkoqi62myJ/LzGmDa5LSwbtqjeyts80O+WI9HN9\n6Xjnf9n6wLT3bdzecnCg688tTaYqI0l/CLyRWnVOvRwdwE19HPt7wCENvx8MPNjH/qwgvaxENghp\n1WOnHLGcDdfetdc8Rb3EmHXW032WaP6YzdVtpxyxnCu/8d2W3XVbxfDVHTtTt1cy2rlTqWN2bjdX\nfO27LRf1KepvZuWWtQ1hDXB4RDw+wGN/AzhM0ipgBjgT+M8D3L8tkiLbUBqnnc4yXXE3MWbtfjg7\nt2e+eqexLnxm1yyf2zzDG190CJtue2i+3aXe5jImLej33hhLq2N3067RanqLB3fNlqrdy8ojaxvC\nPwKvj4ifDvTg0suA/w2MAZdGxIfabV/Wkcr2pMUa6p92nHoDYrNlS8dZuu8+C7aFvUsWX92xk5ld\nswOdBgNY0FC+dHwJQS2JNJucGOf9rzoKSF+ys1utVnpbkbzfegmi3sD7wTXHtNzXsEzhUDWD+twH\nOnWFpM8BxwE3APOlhIh4R9eR9cEJoThZTsxW0wUMam6adseZGB/LVL1Tds29rHo1tkSsXrWMf73n\nkQXVSxPjY7xw5dO4+Z5H9nrNm1evTE0KvU7N4iTSn0FOiTPoqSu+AFwE/CuwueHLRkDW6ZJb1bk/\n+tjcQKdXbjXlQN6D3dJmKV22dHygxxhEMgDYvSe49YEfLUgGAl574gpuuffR1Ndc8bXvpj7eyxQP\nnmK7f0VMrZGpDSEiLpO0L/CLyUN3RUT/HZqtEtqdmFnqvVtt36tWx9kdkVtJQZDarbTTxIBFao4p\nqDVWt6qKatfm0M3jkP2csdaKmFoj61xGvwp8m9rI4o8B30q6n9oIyHpiduqyOKgTudVx6usK1O/g\nB6nVMRvXThC1doBlS8cR5Zye48FkkFaaVo/30kXV8wT1r4iuwVmrjD4MnBYRvxIRLwFOBz6SW1RW\nKllPzE5TYAzqRG43/mDNCSvmVxVbMaDjdZpGofGYWy88jS1/cBrfWf9yPvyG49pO9dGtyYns1VMT\nLcYrHDQ5MT+auFmrx3uZ4sHjHPpXxNQaWRPCeETMV1xFxLeAwVaeWmmlnZiiVi988vobed/G7Zy8\n/kbeeeVW9h9fknoxGuSJ3HxX3mrB+04Jqn6pbmwbgIUrpC1bOt7zvPxrTljBhtcdt6CdYen4EiYn\nxlNXTlu2dJzxplkBRa2x974k2bx59cq2pR+ptv3Fa49teTH54JpjePPqlfMlgjGpZYNy/X1k+bwb\neZ6g/vXyufcray+jS6lVQX4qeehNwD4R8du5RZbCvYyK07j6VKeBURPjY7z2xBV8dcfOwnuYZF0J\nriyy9ubK0nun6F4+RR/fnjTobqf7AW8DfpnaTctNwMcGPFCtIyeE4mWdY6fV3D7DyBc+K7usCSFr\nL6PHgT9JvmyEZW0UfHDX7EhcKLOMjs7jmMP+uVox2iYESZ+JiDdI2k76xHPH5haZldJBkxOZSgiT\nS8dTL5TT9z9SiqqkQemme+UgLuRFJKB+OHlVS6cSwjnJ91fkHYhVQ9qkaM0mxseI2Lsf/Ozcbi6/\n5YH5O4uyX8yyyNq9clAX8m7796ddkOv7WYzpRaqUvKxDL6OIeCj58Xcj4v7GL+B38w/Pyiat50Pz\n6N2L1x7Dj1osxNJczKz6ouZZu1cOatRpN/3700YLn/vZbZz7D9sWZQSxF7Gvnqyznf4G8J6mx16a\n8piNgMYZRltpNdFcmioPVmpVYnrs50+wccvM/Oc0qIFarars0hJT2gV5LmVujLxGEHtwWvW0LSFI\n+h9J+8ERkm5r+PoOsH1xQrQqajV2IU2VByvVS0zNg8aa528a1ECtbvr3d3PhzeMi7cFp1dNpYNrf\nA68EPp98r3+dGBFvyjk2K9DGLTOcvP5GVp23iZPX39h1lUJa1dKbVq/MbbBSv/H2Y80JKzhgv70L\n243VI4MaqNXNYKVuLrx5XKQ9OK162lYZRcSPgB9J+lPgkYj4CYCkp0h6cUR8bTGCtMU1qMbAtKql\nqeceOPAGzTI0XnaqHhnkgjRZquwgvTprfIlALFjBLa+LtBfhqZ6sA9O2AC+MZGNJS4DpiHhhzvEt\n4IFpi6ObxevLoAzxliGGNEX2MrLyGOjANGqJYz5zRMQeSVlfaxVTtcbAMsRb1jWKW5UmnAAsTdbJ\n7e6V9A5J48nXOcC9eQZmxalaY2AZ4i1iIjKzQct6l3828GfA+6h1Jb8BWJdXUFasst7ttlKWeLPW\n7ZuVVda5jB4Gzsw5FiuJqjUGVi1es7Jq26gs6d0R8UeS/pz0uYzekWdwzdyobGbWvUE1Kt+ZfB/o\nVVjS64H3A78EnBQRvsqbNfHEcLbYOo1D+GLy/bIBH/d2YC3wVwPer9lQKMPYChs9naa//iJtFseK\niFf1ctCIuDPZfy8vN+tKFe+0u53VdFRU8W9ZJZ2qjP44+b4WeDbwd8nvZwH35RTTApLWkfRoWrly\n5WIc0oZIVe+0yzC2YlAGdRGv6t+ySjpVGf0zgKSLIuIlDU99UdJN7V4r6XpqSaTZBRHx+awBRsQl\nwCVQa1TO+jobHe0uOFW90+5mVtMyG+RFvKp/yyrJOjBtuaTn1X+RtApY3u4FEXFqRByd8pU5GZh1\nkjbnf+Mso1W90x6WieEGuSZCVf+WVZJ1YNo7gX+SVB+dfCjw33OJaES4LnQwOt01VvVOe1jGVgzy\nIl7Vv2WVZB2Y9mVJhwFHJA/tiIjHez2opNcAf06tlLFJ0taIOL3X/VWN60IHp9MFJ69RzIuR0Idh\n5PMgL+JlGZE+zDJVGUlaCpwLvD0itgErJfW8znJEXB0RB0fEfhHxrFFKBuClBQep0zxGecwx1Kma\nyp40yKovzxeVv6xVRn8LbAb+Q/L794DPAtfkEdSwc13oYGzcMsPPHn9ir8ebLziDvtN242Z2g676\nGoZSU5llTQjPj4g3SjoLICJm5UEEPXNdaP+aq93qli0d58JXHpXrRcMJvTu+iFdH1l5GP5c0QTJI\nTdLzgZ7bEEbdsPQgKVLaXTrA0n33yf3iU4bpts3ykDUhXAh8GThE0uXUpr9+d25RDTnXhfavyLv0\nPBJ6kWtCm9V1rDJKqoZ2UButvBoQcE5E/CDn2Iaai9H9KbLabdD14u51ZmXRMSFEREjaGBEnApsW\nISazjorugjjIhO5GaiuLrFVGt0h6Ua6RmHVhmKrd3EhtZZG1l9EpwNmS7gN+Rq3aKCLi2LwCM+tk\nWKrd3OvMyiJrQnhprlGYjbCiq7/M6jqth7A/cDbwAmA78PGI2HskkJn1bFjmLbLq61RCuAyYA/4v\ntVLCkcA5eQdlNmqGpfrLqq1TQjgyIo4BkPRx4Ov5h2RmZkXo1Mtorv6Dq4rMzIZbpxLCcZJ+nPws\nYCL5vd7L6Km5RmdmZoum0xKaY+2eNzOz4ZF1YJqZmQ25rOMQzKwDL4tqVeeEYNZBlgu9J6izYeAq\nI7M2si6X6WVRbRi4hGDWRtaZSPuZoM5VTVYWhZQQJG2QtEPSbZKuljRZRBxmnWS90Pe6ilrWEojZ\nYiiqyug64OhkttRvAecXFIdZW1kv9L2uojbMVU1eBa56CkkIEfGVhpHPtwAHFxGHWSdZL/S9rs8w\nrGshuORTTWVoQ3grcGXRQZil6WYm0l4mqBvWtRC8Clw15ZYQJF0PPDvlqQsi4vPJNhcATwCXt9nP\nOmAdwMqVK3OI1Ky9PGciHda1EIa15DPscksIEXFqu+clvQV4BfDrERFt9nMJcAnA1NRUy+3MqmhY\n10IY1pLPsCukykjSGcB7gF+JiMeKiMGsLIZxLYRhLfkMu6LaEP4C2A+4ThLALRFxdkGxmA2MxxTU\nDGvJZ9gVkhAi4gVFHNcsT56+YqEqlnxGPaF76gqzARnmMQWjwF1lnRDMBsY9a6rNCd0JwWxgep2+\nwsrBCd0JwWxgep2+wsrBCd0JwWxgep2+wsrBCb0cU1eYDY2y96wZ9V407birrBOC2chwt9jOyp7Q\n8+YqI7MR4V401okTgtmIcC8a68QJwWxEuBeNdeKEYDYi3IvGOnGjstmIcC8a68QJwWyEjHovGmvP\nVUZmZga4hGAjygO0zPbmhGAjxwO0zNK5yshGjgdomaVzQrCR4wFaZumcEGzkeICWWTonBBs5HqBl\nls6NyjZyPEDLLF0hCUHSRcCrgT3Aw8BvRcSDRcRio8kDtMz2VlSV0YaIODYijgeuAf6goDjMzCxR\nSEKIiB83/HoAEEXEYWZmTyqsDUHSh4D/AvwIOKXNduuAdQArV65cnODMzEaQIvK5OZd0PfDslKcu\niIjPN2x3PrB/RFzYaZ9TU1MxPT09wCjNzIafpM0RMdVpu9xKCBFxasZN/x7YBHRMCGZmlp9C2hAk\nHdbw66uAHUXEYWZmTyqqDWG9pMOpdTu9Hzi7oDjMzCxRSEKIiNcWcVwzM2vNU1eYmRngqSusgry4\njVk+nBCsUry4jVl+XGVkleLFbczy44RgleLFbczy44RgleLFbczy44RgleLFbczy40ZlqxQvbmOW\nHycEqxwvbmOWD1cZmZkZ4IRgZmYJJwQzMwOcEMzMLOGEYGZmgBOCmZklnBDMzAxwQjAzs4QTgpmZ\nAU4IZmaWcEIwMzPACcHMzBKFTm4n6feBDcDyiPhBHsfw+rtmZtkUlhAkHQL8BvBAXsfw+rtmZtkV\nWWX0EeDdQOR1AK+/a2aWXSEJQdKrgJmI2JZh23WSpiVN79y5s6vjeP1dM7PscqsyknQ98OyUpy4A\n3guclmU/EXEJcAnA1NRUV6WJgyYnmEm5+Hv9XTOzveVWQoiIUyPi6OYv4F5gFbBN0n3AwcCtktKS\nR1+8/q6ZWXaL3qgcEduBZ9Z/T5LCVB69jLz+rplZdkO/prLX3zUzy6bwhBARhxYdg5mZeaSymZkl\nnBDMzAxwQjAzs4QTgpmZAaCI3GaOGDhJO4H7+9jFM4BcJtEbEMfXH8fXn7LHB+WPsazxPTcilnfa\nqFIJoV+SpiNiqug4WnF8/XF8/Sl7fFD+GMseXyeuMjIzM8AJwczMEqOWEC4pOoAOHF9/HF9/yh4f\nlD/GssfX1ki1IZiZWWujVkIwM7MWnBDMzAwYkYQgaYOkHZJuk3S1pMmG586XdLekuySdXlB8r5d0\nh6Q9kqYaHj9U0qykrcnXX5YpvuS5wj+/pnjeL2mm4TN7WdExAUg6I/mM7pZ0XtHxNJN0n6TtyWc2\nXYJ4LpX0sKTbGx47UNJ1kr6dfF9WsvhKee51YyQSAnAdcHREHAt8CzgfQNKRwJnAUcAZwMckjbXc\nS35uB9YCN6U8d09EHJ98nb3IcdWlxleiz6/ZRxo+sy8VHUzymXwUeClwJHBW8tmVzSnJZ1aGfvSf\noHZONToPuCEiDgNuSH4vyifYOz4o2bnXrZFICBHxlYh4Ivn1FmqrtAG8Gvh0RDweEd8B7gZOKiC+\nOyPirsU+blZt4ivF51cBJwF3R8S9EfFz4NPUPjtrISJuAh5pevjVwGXJz5cBaxY1qAYt4qu8kUgI\nTd4K/GPy8wrguw3PfS95rExWSdoi6Z8l/aeig2lS1s/v7Un14KVFVis0KOvn1CiAr0jaLGld0cG0\n8KyIeAgg+f7MDtsXoWznXlcKXyBnUCRdD6Sty3xBRHw+2eYC4Ang8vrLUrbPpR9ulvhSPASsjIgf\nSjoR2CjpqIj4cUniW7TPb8FB28QK/B/goiSOi4APU7sJKFIhn1OXTo6IByU9E7hO0o7kLtiyK+O5\n15WhSQgRcWq75yW9BXgF8Ovx5OCL7wGHNGx2MPBgEfG1eM3jwOPJz5sl3QP8IjDwRr9e4mMRP79G\nWWOV9NfANTmHk0Uhn1M3IuLB5PvDkq6mVs1VtoTwfUnPiYiHJD0HeLjogBpFxPfrP5fo3OvKSFQZ\nSToDeA/wqoh4rOGpLwBnStpP0irgMODrRcSYRtLyeiOtpOdRi+/eYqNaoHSfX3KhqHsNtQbxon0D\nOEzSKkn7UmuI/0LBMc2TdICkp9R/Bk6jHJ9bsy8Ab0l+fgvQquRaiJKee92JiKH/otbY+V1ga/L1\nlw3PXQDcA9wFvLSg+F5D7S7yceD7wLXJ468F7gC2AbcCryxTfGX5/Jpi/RSwHbiN2gXkOUXHlMT1\nMmo93O6hVg1XeEwNsT0vOce2Jedb4fEBV1CrMp1Lzr3fAZ5OrXfRt5PvB5YsvlKee918eeoKMzMD\nRqTKyMzMOnNCMDMzwAnBzMwSTghmZgY4IZiZWcIJwYaOpKc3zDj5/5pmoNx3QMd4iqQfSvqFpsev\nkbS2zetOlbRxEDGYDdrQjFQ2q4uIHwLHQ21KYuCnEfHHjdtIErUVA/f0eIyfSLqR2oRrlyf7XAa8\nGHhd79GbFcclBBsZkl4g6fZkXYlbgUMk7Wp4/kxJf5P8/CxJV0malvR1SatTdnkFtVHHda8FNkXE\nv0taLenfkokJb5Z0WEo8H5T0ew2/75B0cPLzW5LjbpX0MUlLJO0j6VPJugW3S3rHYD4ZsxonBBs1\nRwIfj4gTgJk22/0Z8EdRWxvgDcDfpGyzCVjdMKvlmdSSBMCdwC8nx7kI+GDWACUdTW10+H+MiOOp\nleTPBE4EnhERx0TE0cAns+7TLAtXGdmouScivpFhu1OBw2s1SwAskzQREbP1ByLicUmbgLWSrqG2\nUNANydOTwCclPb+HGE8FXgRMJ8efoDb1yrVJTH8KfAn4Sg/7NmvJCcFGzc8aft7Dwqmp92/4WcBJ\nUVvQpp0rgN+ndtG+Kp5ciOlD1OZ8+pikFwBfTnntEywspdePL+DSiPhfzS+QdCy1ldfeQa2Kqqxr\nF1gFucrIRlbSoPyopMMkLaFWTVN3PfC2+i+Sjm+xm+uplQzO5snqIoCn8WSV1G+1eO191KqBkHQS\nT06RfT3wBknPSJ57uqSVkpZTawj/LHAh8MIMb9MsMycEG3XvoXb3fgO1WSvr3gacnKx+9U3gv6W9\nOCJ2A1cDTwVubnjqD4ENkm5Oe13is8CzJG2hNlvmvck+twMfAK6XdBu1qqFnUUsYN0naCvw18N4u\n36tZW57t1MzMAJcQzMws4YSCUrYtAAAAJElEQVRgZmaAE4KZmSWcEMzMDHBCMDOzhBOCmZkBTghm\nZpb4//DfxTJNKt5jAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11fae0c18>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "xgb.fit(X_train, y_train)\n",
    "predictions = xgb.predict(X_test)\n",
    "plt.scatter(y_test, predictions)\n",
    "plt.xlabel(\"True Values\")\n",
    "plt.ylabel(\"Predictions\")\n",
    "print(\"Score:\", xgb.score(X_test, y_test))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Well, it's positive."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Let's turn this into a classification problem\n",
    "\n",
    "(And pray)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock['target_class'] = stock.target.apply(lambda x: 1 if x >= 0 else 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    957\n",
       "1    909\n",
       "Name: target_class, dtype: int64"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stock.target_class.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(stock.iloc[915:-1].drop(['target', 'target_class'], 1), stock.iloc[915:-1]['target_class'], test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluation Function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "def eval_sklearn_model(y_true, predictions, model=None, X=None):\n",
    "    \"\"\"This function takes the true values for y and the predictions made by the model and prints out the confusion matrix along with Accuracy, Precision, and, if model and X provided, Roc_Auc Scores.\"\"\"\n",
    "    cnf_matrix = confusion_matrix(y_true, predictions)\n",
    "\n",
    "    print('True Negative: ', cnf_matrix[0, 0], '| False Positive: ', cnf_matrix[0, 1])\n",
    "    print('False Negative: ', cnf_matrix[1, 0], '| True Positive: ', cnf_matrix[1, 1], '\\n')\n",
    "\n",
    "    sensitivity = cnf_matrix[1, 1]/ (cnf_matrix[1, 0] + cnf_matrix[1, 1])\n",
    "    specificity = cnf_matrix[0, 0]/ (cnf_matrix[0, 1] + cnf_matrix[0, 0])\n",
    "\n",
    "    print('Sensitivity (TP/ TP + FN): ', sensitivity)\n",
    "    print('Specificity (TN/ TN + FP): ', specificity, '\\n')\n",
    "\n",
    "    print('Accuracy: ', accuracy_score(y_true, predictions, normalize=True))\n",
    "    print('Precision: ', precision_score(y_true, predictions))\n",
    "    if model != None:\n",
    "        print('Roc-Auc: ', roc_auc_score(y_true, [x[1] for x in model.predict_proba(X)]))\n",
    "    else:\n",
    "        pass\n",
    "    print('\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "from xgboost import XGBClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True Negative:  87 | False Positive:  53\n",
      "False Negative:  92 | True Positive:  53 \n",
      "\n",
      "Sensitivity (TP/ TP + FN):  0.365517241379\n",
      "Specificity (TN/ TN + FP):  0.621428571429 \n",
      "\n",
      "Accuracy:  0.491228070175\n",
      "Precision:  0.5\n",
      "Roc-Auc:  0.510123152709\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "xgb = XGBClassifier(objective='binary:logistic')\n",
    "xgb.fit(X_train, y_train)\n",
    "test_predictions = xgb.predict(X_test)\n",
    "eval_sklearn_model(y_test, test_predictions, model=xgb, X=X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sentiment Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/anaconda/lib/python3.6/site-packages/nltk/twitter/__init__.py:20: UserWarning: The twython library has not been installed. Some functionality from the twitter package will not be available.\n",
      "  warnings.warn(\"The twython library has not been installed. \"\n"
     ]
    }
   ],
   "source": [
    "from nltk.sentiment.vader import SentimentIntensityAnalyzer as SIA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reload data as I dropped the nlp columns with count vect\n",
    "stock = pd.read_csv('TESLAHistoricalQuotes.csv')\n",
    "stock = stock.drop(0,0)\n",
    "stock.date = pd.to_datetime(stock.date)\n",
    "stock.date = stock.date.apply(lambda x: x.date())\n",
    "stock = stock.sort_values('date')\n",
    "stock.volume = stock.volume.astype(float)\n",
    "stock.set_index(stock.date, inplace=True)\n",
    "stock = stock.drop('date', 1)\n",
    "stock = pd.merge(stock, techmeme, how='left', left_index=True, right_index=True)\n",
    "stock = pd.merge(stock, tesla, how='left', left_index=True, right_index=True)\n",
    "stock = pd.merge(stock, elon, how='left', left_index=True, right_index=True)\n",
    "stock['daily_gain'] = stock.close - stock.open\n",
    "stock['daily_change'] = stock.daily_gain / stock.open\n",
    "stock['target'] = stock.daily_gain.shift(-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock.news_sources = stock.news_sources.fillna('')\n",
    "stock.news_text = stock.news_text.fillna('')\n",
    "stock.tesla_tweet = stock.tesla_tweet.fillna('')\n",
    "stock.elon_tweet = stock.elon_tweet.fillna('')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "sia = SIA()\n",
    "sentiment = stock['news_text'].apply(sia.polarity_scores)\n",
    "sent = pd.DataFrame(list(sentiment))\n",
    "sent.index = stock.index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "sent.columns = ['news_compound', 'news_neg', 'news_neu', 'news_pos']\n",
    "stock = pd.merge(stock, sent, how='left', left_index=True, right_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentiment = stock['elon_tweet'].apply(sia.polarity_scores)\n",
    "sent = pd.DataFrame(list(sentiment))\n",
    "sent.index = stock.index\n",
    "sent.columns = ['elon_compound', 'elon_neg', 'elon_neu', 'elon_pos']\n",
    "stock = pd.merge(stock, sent, how='left', left_index=True, right_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "sentiment = stock['tesla_tweet'].apply(sia.polarity_scores)\n",
    "sent = pd.DataFrame(list(sentiment))\n",
    "sent.index = stock.index\n",
    "sent.columns = ['telsa_compound', 'tesla_neg', 'tesla_neu', 'tesla_pos']\n",
    "stock = pd.merge(stock, sent, how='left', left_index=True, right_index=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Note on earliest NLP relevant info\n",
    "1. elon earliest tweet 2013-03-23\n",
    "2. tesla earliest tweet 2014-02-18\n",
    "3. techmeme earliest story 2006-12-07\n",
    "\n",
    "```stock.iloc[915]``` is 2014-02-18"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(951, 24)"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stock.iloc[915:].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['close', 'volume', 'open', 'high', 'low', 'news_sources', 'news_text',\n",
       "       'tesla_tweet', 'elon_tweet', 'daily_gain', 'daily_change', 'target',\n",
       "       'news_compound', 'news_neg', 'news_neu', 'news_pos', 'elon_compound',\n",
       "       'elon_neg', 'elon_neu', 'elon_pos', 'telsa_compound', 'tesla_neg',\n",
       "       'tesla_neu', 'tesla_pos'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stock.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = stock.iloc[915:-1].drop(['target', 'news_sources', 'news_text', 'tesla_tweet', 'elon_tweet'], 1)\n",
    "y = stock.iloc[915:-1].target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "XGBRegressor(base_score=0.5, colsample_bylevel=1, colsample_bytree=1, gamma=0,\n",
       "       learning_rate=0.02, max_delta_step=0, max_depth=3,\n",
       "       min_child_weight=1, missing=None, n_estimators=100, nthread=-1,\n",
       "       objective='reg:linear', reg_alpha=0, reg_lambda=1,\n",
       "       scale_pos_weight=1, seed=0, silent=True, subsample=1)"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xgb = XGBRegressor(gamma=0, learning_rate=0.02, max_depth=3, n_estimators=100)\n",
    "xgb.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 Accuracy: 0.143272882714\n"
     ]
    }
   ],
   "source": [
    "accuracy = metrics.r2_score(y_train, xgb.predict(X_train))\n",
    "print(\"R2 Accuracy:\", accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Score: 0.0103119613995\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYQAAAEKCAYAAAASByJ7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAZQQREQkooQgIiIA/H87iq7emMdIEwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x110934630>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "predictions = xgb.predict(X_test)\n",
    "plt.scatter(y_test, predictions)\n",
    "plt.xlabel(\"True Values\")\n",
    "plt.ylabel(\"Predictions\")\n",
    "print(\"Score:\", xgb.score(X_test, y_test))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Well... I mean.  Yup."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# To Classification!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock['target_class'] = stock.target.apply(lambda x: 1 if x >= 0 else 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = stock.iloc[915:-1].drop(['target', 'target_class', 'news_sources', 'news_text', 'tesla_tweet', 'elon_tweet'], 1)\n",
    "y = stock.iloc[915:-1].target_class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True Negative:  75 | False Positive:  65\n",
      "False Negative:  80 | True Positive:  65 \n",
      "\n",
      "Sensitivity (TP/ TP + FN):  0.448275862069\n",
      "Specificity (TN/ TN + FP):  0.535714285714 \n",
      "\n",
      "Accuracy:  0.491228070175\n",
      "Precision:  0.5\n",
      "Roc-Auc:  0.525738916256\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "xgb = XGBClassifier(objective='binary:logistic')\n",
    "xgb.fit(X_train, y_train)\n",
    "test_predictions = xgb.predict(X_test)\n",
    "eval_sklearn_model(y_test, test_predictions, model=xgb, X=X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Grumble, grumble"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Can I predict the spread of a Day?\n",
    "\n",
    "Will volatility be changed by the news?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "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>close</th>\n",
       "      <th>volume</th>\n",
       "      <th>open</th>\n",
       "      <th>high</th>\n",
       "      <th>low</th>\n",
       "      <th>news_sources</th>\n",
       "      <th>news_text</th>\n",
       "      <th>tesla_tweet</th>\n",
       "      <th>elon_tweet</th>\n",
       "      <th>daily_gain</th>\n",
       "      <th>...</th>\n",
       "      <th>news_pos</th>\n",
       "      <th>elon_compound</th>\n",
       "      <th>elon_neg</th>\n",
       "      <th>elon_neu</th>\n",
       "      <th>elon_pos</th>\n",
       "      <th>telsa_compound</th>\n",
       "      <th>tesla_neg</th>\n",
       "      <th>tesla_neu</th>\n",
       "      <th>tesla_pos</th>\n",
       "      <th>target_class</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2010-06-29</th>\n",
       "      <td>23.89</td>\n",
       "      <td>18751150.0</td>\n",
       "      <td>19.00</td>\n",
       "      <td>25.0000</td>\n",
       "      <td>17.54</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>4.89</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-06-30</th>\n",
       "      <td>23.83</td>\n",
       "      <td>17165210.0</td>\n",
       "      <td>25.79</td>\n",
       "      <td>30.4192</td>\n",
       "      <td>23.30</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>-1.96</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-07-01</th>\n",
       "      <td>21.96</td>\n",
       "      <td>8216789.0</td>\n",
       "      <td>25.00</td>\n",
       "      <td>25.9200</td>\n",
       "      <td>20.27</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>-3.04</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-07-02</th>\n",
       "      <td>19.20</td>\n",
       "      <td>5135795.0</td>\n",
       "      <td>23.00</td>\n",
       "      <td>23.1000</td>\n",
       "      <td>18.71</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>-3.80</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-07-06</th>\n",
       "      <td>16.11</td>\n",
       "      <td>6858092.0</td>\n",
       "      <td>20.00</td>\n",
       "      <td>20.0000</td>\n",
       "      <td>15.83</td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td></td>\n",
       "      <td>-3.89</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 25 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            close      volume   open     high    low news_sources news_text  \\\n",
       "date                                                                          \n",
       "2010-06-29  23.89  18751150.0  19.00  25.0000  17.54                          \n",
       "2010-06-30  23.83  17165210.0  25.79  30.4192  23.30                          \n",
       "2010-07-01  21.96   8216789.0  25.00  25.9200  20.27                          \n",
       "2010-07-02  19.20   5135795.0  23.00  23.1000  18.71                          \n",
       "2010-07-06  16.11   6858092.0  20.00  20.0000  15.83                          \n",
       "\n",
       "           tesla_tweet elon_tweet  daily_gain      ...       news_pos  \\\n",
       "date                                               ...                  \n",
       "2010-06-29                               4.89      ...            0.0   \n",
       "2010-06-30                              -1.96      ...            0.0   \n",
       "2010-07-01                              -3.04      ...            0.0   \n",
       "2010-07-02                              -3.80      ...            0.0   \n",
       "2010-07-06                              -3.89      ...            0.0   \n",
       "\n",
       "            elon_compound  elon_neg  elon_neu  elon_pos  telsa_compound  \\\n",
       "date                                                                      \n",
       "2010-06-29            0.0       0.0       0.0       0.0             0.0   \n",
       "2010-06-30            0.0       0.0       0.0       0.0             0.0   \n",
       "2010-07-01            0.0       0.0       0.0       0.0             0.0   \n",
       "2010-07-02            0.0       0.0       0.0       0.0             0.0   \n",
       "2010-07-06            0.0       0.0       0.0       0.0             0.0   \n",
       "\n",
       "            tesla_neg  tesla_neu  tesla_pos  target_class  \n",
       "date                                                       \n",
       "2010-06-29        0.0        0.0        0.0             0  \n",
       "2010-06-30        0.0        0.0        0.0             0  \n",
       "2010-07-01        0.0        0.0        0.0             0  \n",
       "2010-07-02        0.0        0.0        0.0             0  \n",
       "2010-07-06        0.0        0.0        0.0             0  \n",
       "\n",
       "[5 rows x 25 columns]"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stock.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock['volatility'] = stock.high - stock.low"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = stock.iloc[915:-1].drop(['target', 'target_class', 'volatility', 'high', 'low', 'close', 'volume', 'open', 'news_sources', 'news_text', 'tesla_tweet', 'elon_tweet', 'daily_gain', 'daily_change'], 1)\n",
    "y = stock.iloc[915:-1].volatility"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 Accuracy: 0.0287532757729\n"
     ]
    }
   ],
   "source": [
    "xgb = XGBRegressor(gamma=0, learning_rate=0.02, max_depth=3, n_estimators=100)\n",
    "xgb.fit(X_train, y_train)\n",
    "accuracy = metrics.r2_score(y_train, xgb.predict(X_train))\n",
    "print(\"R2 Accuracy:\", accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Score: -0.0883789185676\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEKCAYAAAAfGVI8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFAhERApOgUBEpOAUCERECk6BQESk4P4/jr61MmnKvK0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1149bedd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "predictions = xgb.predict(X_test)\n",
    "plt.scatter(y_test, predictions)\n",
    "plt.xlabel(\"True Values\")\n",
    "plt.ylabel(\"Predictions\")\n",
    "print(\"Score:\", xgb.score(X_test, y_test))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## With Count Vectorizer?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sent_analysis(dataframe):\n",
    "    sia = SIA()\n",
    "    sentiment = dataframe['news_text'].apply(sia.polarity_scores)\n",
    "    sent = pd.DataFrame(list(sentiment))\n",
    "    sent.index = dataframe.index\n",
    "    \n",
    "    sent.columns = ['news_compound', 'news_neg', 'news_neu', 'news_pos']\n",
    "    dataframe = pd.merge(dataframe, sent, how='left', left_index=True, right_index=True)\n",
    "    \n",
    "    sentiment = dataframe['elon_tweet'].apply(sia.polarity_scores)\n",
    "    sent = pd.DataFrame(list(sentiment))\n",
    "    sent.index = dataframe.index\n",
    "    sent.columns = ['elon_compound', 'elon_neg', 'elon_neu', 'elon_pos']\n",
    "    dataframe = pd.merge(dataframe, sent, how='left', left_index=True, right_index=True)\n",
    "\n",
    "    sentiment = dataframe['tesla_tweet'].apply(sia.polarity_scores)\n",
    "    sent = pd.DataFrame(list(sentiment))\n",
    "    sent.index = dataframe.index\n",
    "    sent.columns = ['telsa_compound', 'tesla_neg', 'tesla_neu', 'tesla_pos']\n",
    "    dataframe = pd.merge(dataframe, sent, how='left', left_index=True, right_index=True)\n",
    "    \n",
    "    return dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock = pd.read_csv('TESLAHistoricalQuotes.csv')\n",
    "stock = stock.drop(0,0)\n",
    "stock.date = pd.to_datetime(stock.date)\n",
    "stock.date = stock.date.apply(lambda x: x.date())\n",
    "stock = stock.sort_values('date')\n",
    "stock.volume = stock.volume.astype(float)\n",
    "stock.set_index(stock.date, inplace=True)\n",
    "stock = stock.drop('date', 1)\n",
    "stock = pd.merge(stock, techmeme, how='left', left_index=True, right_index=True)\n",
    "stock = pd.merge(stock, tesla, how='left', left_index=True, right_index=True)\n",
    "stock = pd.merge(stock, elon, how='left', left_index=True, right_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock.news_text = stock.news_text.fillna('')\n",
    "stock.tesla_tweet = stock.tesla_tweet.fillna('')\n",
    "stock.elon_tweet = stock.elon_tweet.fillna('')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 214,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock = sent_analysis(stock)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 215,
   "metadata": {},
   "outputs": [],
   "source": [
    "temp_df = stock[stock.news_text.notnull()]\n",
    "cvec = CountVectorizer(stop_words=stopwords.words('english'), ngram_range=(1,2), max_features=1000)\n",
    "cvec.fit(temp_df.news_text.dropna())\n",
    "cvec_df  = pd.DataFrame(cvec.transform(temp_df.news_text.dropna()).todense(), columns=cvec.get_feature_names())\n",
    "cvec_df.index = temp_df.index\n",
    "cvec_df.columns += \" _news_\"\n",
    "stock = stock.join(cvec_df).drop('news_text',1)\n",
    "\n",
    "temp_df = stock[stock.tesla_tweet.notnull()]\n",
    "cvec.fit(temp_df.tesla_tweet.dropna())\n",
    "cvec_df  = pd.DataFrame(cvec.transform(temp_df.tesla_tweet.dropna()).todense(), columns=cvec.get_feature_names())\n",
    "cvec_df.index = temp_df.index\n",
    "cvec_df.columns += \" _tesla_\"\n",
    "stock = stock.join(cvec_df).drop('tesla_tweet',1)\n",
    "\n",
    "temp_df = stock[stock.elon_tweet.notnull()]\n",
    "cvec.fit(temp_df.elon_tweet.dropna())\n",
    "cvec_df  = pd.DataFrame(cvec.transform(temp_df.elon_tweet.dropna()).todense(), columns=cvec.get_feature_names())\n",
    "cvec_df.index = temp_df.index\n",
    "cvec_df.columns += \" _elon_\"\n",
    "stock = stock.join(cvec_df).drop('elon_tweet',1)\n",
    "\n",
    "del temp_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 216,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock.news_sources = stock.news_sources.fillna('')\n",
    "mlb = MultiLabelBinarizer()\n",
    "stock = stock.join(pd.DataFrame(mlb.fit_transform(stock.news_sources), columns=mlb.classes_, index=stock.index))\n",
    "stock = stock.drop('news_sources', 1)\n",
    "stock = stock.fillna(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 217,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock['volatility'] = stock.high - stock.low"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 222,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = stock.iloc[925:-1].drop(['volatility', 'volume', 'high', 'low', 'close', 'open'], 1)\n",
    "y = stock.iloc[925:-1].volatility"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 223,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 224,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 Accuracy: 0.0503138202204\n"
     ]
    }
   ],
   "source": [
    "xgb = XGBRegressor(gamma=0, learning_rate=0.02, max_depth=3, n_estimators=100)\n",
    "xgb.fit(X_train, y_train)\n",
    "accuracy = metrics.r2_score(y_train, xgb.predict(X_train))\n",
    "print(\"R2 Accuracy:\", accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Score: -0.080712113206\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEKCAYAAAAfGVI8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1107123c8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "predictions = xgb.predict(X_test)\n",
    "plt.scatter(y_test, predictions)\n",
    "plt.xlabel(\"True Values\")\n",
    "plt.ylabel(\"Predictions\")\n",
    "print(\"Score:\", xgb.score(X_test, y_test))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Daily Change?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock['daily_gain'] = stock.close - stock.open\n",
    "stock['daily_change'] = stock.daily_gain / stock.open"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "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>close</th>\n",
       "      <th>volume</th>\n",
       "      <th>open</th>\n",
       "      <th>high</th>\n",
       "      <th>low</th>\n",
       "      <th>news_compound</th>\n",
       "      <th>news_neg</th>\n",
       "      <th>news_neu</th>\n",
       "      <th>news_pos</th>\n",
       "      <th>elon_compound</th>\n",
       "      <th>...</th>\n",
       "      <th>iLounge</th>\n",
       "      <th>iMore</th>\n",
       "      <th>iPhone Hacks</th>\n",
       "      <th>iPhone in Canada Blog</th>\n",
       "      <th>lists.swift.org Mailing Lists</th>\n",
       "      <th>volatility</th>\n",
       "      <th>next_day_open</th>\n",
       "      <th>daily_gain</th>\n",
       "      <th>daily_change</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2010-06-29</th>\n",
       "      <td>23.89</td>\n",
       "      <td>18751150.0</td>\n",
       "      <td>19.00</td>\n",
       "      <td>25.0000</td>\n",
       "      <td>17.54</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7.4600</td>\n",
       "      <td>25.79</td>\n",
       "      <td>4.89</td>\n",
       "      <td>0.257368</td>\n",
       "      <td>-0.075998</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-06-30</th>\n",
       "      <td>23.83</td>\n",
       "      <td>17165210.0</td>\n",
       "      <td>25.79</td>\n",
       "      <td>30.4192</td>\n",
       "      <td>23.30</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>7.1192</td>\n",
       "      <td>25.00</td>\n",
       "      <td>-1.96</td>\n",
       "      <td>-0.075998</td>\n",
       "      <td>-0.121600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-07-01</th>\n",
       "      <td>21.96</td>\n",
       "      <td>8216789.0</td>\n",
       "      <td>25.00</td>\n",
       "      <td>25.9200</td>\n",
       "      <td>20.27</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>5.6500</td>\n",
       "      <td>23.00</td>\n",
       "      <td>-3.04</td>\n",
       "      <td>-0.121600</td>\n",
       "      <td>-0.165217</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-07-02</th>\n",
       "      <td>19.20</td>\n",
       "      <td>5135795.0</td>\n",
       "      <td>23.00</td>\n",
       "      <td>23.1000</td>\n",
       "      <td>18.71</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4.3900</td>\n",
       "      <td>20.00</td>\n",
       "      <td>-3.80</td>\n",
       "      <td>-0.165217</td>\n",
       "      <td>-0.194500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-07-06</th>\n",
       "      <td>16.11</td>\n",
       "      <td>6858092.0</td>\n",
       "      <td>20.00</td>\n",
       "      <td>20.0000</td>\n",
       "      <td>15.83</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>4.1700</td>\n",
       "      <td>16.40</td>\n",
       "      <td>-3.89</td>\n",
       "      <td>-0.194500</td>\n",
       "      <td>-0.036585</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 3320 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            close      volume   open     high    low  news_compound  news_neg  \\\n",
       "date                                                                            \n",
       "2010-06-29  23.89  18751150.0  19.00  25.0000  17.54            0.0       0.0   \n",
       "2010-06-30  23.83  17165210.0  25.79  30.4192  23.30            0.0       0.0   \n",
       "2010-07-01  21.96   8216789.0  25.00  25.9200  20.27            0.0       0.0   \n",
       "2010-07-02  19.20   5135795.0  23.00  23.1000  18.71            0.0       0.0   \n",
       "2010-07-06  16.11   6858092.0  20.00  20.0000  15.83            0.0       0.0   \n",
       "\n",
       "            news_neu  news_pos  elon_compound    ...     iLounge  iMore  \\\n",
       "date                                             ...                      \n",
       "2010-06-29       0.0       0.0            0.0    ...           0      0   \n",
       "2010-06-30       0.0       0.0            0.0    ...           0      0   \n",
       "2010-07-01       0.0       0.0            0.0    ...           0      0   \n",
       "2010-07-02       0.0       0.0            0.0    ...           0      0   \n",
       "2010-07-06       0.0       0.0            0.0    ...           0      0   \n",
       "\n",
       "            iPhone Hacks  iPhone in Canada Blog  \\\n",
       "date                                              \n",
       "2010-06-29             0                      0   \n",
       "2010-06-30             0                      0   \n",
       "2010-07-01             0                      0   \n",
       "2010-07-02             0                      0   \n",
       "2010-07-06             0                      0   \n",
       "\n",
       "            lists.swift.org Mailing Lists  volatility  next_day_open  \\\n",
       "date                                                                   \n",
       "2010-06-29                              0      7.4600          25.79   \n",
       "2010-06-30                              0      7.1192          25.00   \n",
       "2010-07-01                              0      5.6500          23.00   \n",
       "2010-07-02                              0      4.3900          20.00   \n",
       "2010-07-06                              0      4.1700          16.40   \n",
       "\n",
       "            daily_gain  daily_change    target  \n",
       "date                                            \n",
       "2010-06-29        4.89      0.257368 -0.075998  \n",
       "2010-06-30       -1.96     -0.075998 -0.121600  \n",
       "2010-07-01       -3.04     -0.121600 -0.165217  \n",
       "2010-07-02       -3.80     -0.165217 -0.194500  \n",
       "2010-07-06       -3.89     -0.194500 -0.036585  \n",
       "\n",
       "[5 rows x 3320 columns]"
      ]
     },
     "execution_count": 200,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "stock['target'] = stock.daily_change.shift(-1)\n",
    "stock.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {},
   "outputs": [],
   "source": [
    "stock['next_day_open'] = stock.open.shift(-1) \n",
    "stock['true_grow'] = stock.target.apply(lambda x: 1 if x >= 0 else 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = stock.iloc[925:-1].drop(['volatility', 'volume', 'high', 'low', 'close', 'open', 'target', 'true_grow'], 1)\n",
    "y = stock.iloc[925:-1].true_grow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True Negative:  69 | False Positive:  73\n",
      "False Negative:  68 | True Positive:  72 \n",
      "\n",
      "Sensitivity (TP/ TP + FN):  0.514285714286\n",
      "Specificity (TN/ TN + FP):  0.485915492958 \n",
      "\n",
      "Accuracy:  0.5\n",
      "Precision:  0.496551724138\n",
      "Roc-Auc:  0.490945674044\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "xgb = XGBClassifier(objective='binary:logistic')\n",
    "xgb.fit(X_train, y_train)\n",
    "test_predictions = xgb.predict(X_test)\n",
    "eval_sklearn_model(y_test, test_predictions, model=xgb, X=X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}