Iron_Condor_v06_stock-price_rolling-fit -and-forecast.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Notebook for training for stock price with rolling fit and rolling forecast\n",
    "Forecast for last 10% of data. Fit is done in steps of 5 days, forecast is done day by day.\n",
    "- Sequential model\n",
    "- LSTM model\n",
    "- not stationary\n",
    "- rolling fit\n",
    "- rolling forecast"
   ]
  },
  {
   "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 matplotlib.pyplot as plt\n",
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "from matplotlib import cm\n",
    "import numpy as np\n",
    "import scipy.stats as si\n",
    "from IPython.display import Image\n",
    "import datetime\n",
    "import time\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Activation, Dropout\n",
    "from keras.layers import LSTM\n",
    "from sklearn import preprocessing, metrics\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "backup = pd.HDFStore('backup_v06.h5')\n",
    "#df = backup['df']\n",
    "#unscaled_seq_predictions = backup['unscaled_seq_predictions']\n",
    "#unscaled_data = backup['unscaled_data']\n",
    "#unscaled_lstm_predictions = backup['unscaled_lstm_predictions']\n",
    "#test_data_unscaled = backup['test_data_unscaled']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Iron Condor\n",
    "The iron condor is a investment strategy using four option with the same strike date. More precise, the iron condor uses two vertical spreads, one put spread and a call spread. The put spread consists at of a short put option (P_s) with a strike price below the actual stock price and a long put option (P_l) with a strike price below the short put option P_s. The call spread consists of a short call option (C_s) above the actual stock price and a long call option (C_l) above the short call option C_s. All options have the same strike date and the spread on both legs can vary but is chosen mostly the same.\n",
    "![title](Condor_strategy.png)\n",
    "\"A trader who buys an iron condor speculates that the spot price of the underlying instrument will be between the short strikes when the options expire where the position is the most profitable. Thus, the iron condor is an options strategy considered when the trader has a neutral outlook for the market. Buying iron condors are popular with traders who seek regular income from their trading capital. An iron condor buyer will attempt to construct the trade so that the short strikes are close enough that the position will earn a desirable net credit, but wide enough apart so that it is likely that the spot price of the underlying will remain between the short strikes for the duration of the options contract. The trader would typically play iron condors every month (if possible) thus generating monthly income with the strategy.\" [https://en.wikipedia.org/wiki/Iron_condor]\n",
    "\n",
    "### Goal of the project\n",
    "The goal of the project is to find a strategy to maximise the profit with an iron condor. The ideal short and long spreads should be found at any time based on data of the underlying stock and of the indices S&P500 and Nasdaq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_data(var_name,name):\n",
    "    # reads in csv into DataFrame, keeps Adj Close and Volume and calculates rolling averages and \n",
    "    # rolling standard deviation of Adj Close for 4, 9 and 18 days\n",
    "    # var_name: path to csv file\n",
    "    # name: name of the column df\n",
    "    df = pd.read_csv('data/'+var_name+'.csv',index_col='Date', parse_dates=True)\n",
    "    df.rename(columns={\"Adj Close\": name, \"Volume\": name+'_volume', \"Open\": name+'_open'}, inplace = True)\n",
    "    df.rename(columns={\"High\": name+'_high', \"Low\": name+'_low', \"Close\": name+'_close'}, inplace = True)\n",
    "    #df.drop(['Open','High','Low','Close'],axis=1, inplace = True)\n",
    "    #if name == 'google':\n",
    "        #df['google_45d'] = df['google'].shift(-32) # stock price in 45 days (approx. 32 trading days)\n",
    "    df[name+'_ra_04'] = df[name].rolling('4d').mean()\n",
    "    df[name+'_std_04'] = df[name].rolling('4d').std()\n",
    "    df[name+'_ra_09'] = df[name].rolling('9d').mean()\n",
    "    df[name+'_std_09'] = df[name].rolling('9d').std()\n",
    "    df[name+'_ra_18'] = df[name].rolling('18d').mean()\n",
    "    df[name+'_std_18'] = df[name].rolling('18d').std()\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sigma_fct(df, name, days = 10):\n",
    "    # For Black-Scholes I need the volatility of return of the underlying assets in the last days\n",
    "    # based on https://www.wikihow.com/Calculate-Historical-Stock-Volatility\n",
    "    # First the ln of daily returns is calculated for each element of names\n",
    "    for name in names:\n",
    "        a=pd.DataFrame()\n",
    "        a['ln_daily_return'] = np.log(df[name] / df[name].shift(1))\n",
    "        # next the rolling average (mean) over certain days is calculated\n",
    "        a['rol_avg'] =  a['ln_daily_return'].rolling(str(days)+'d').mean()\n",
    "        # next deviation from the mean is calculated\n",
    "        a['dev_mean'] = a['ln_daily_return'] - a['rol_avg']\n",
    "        # next variance is calculated for certain days\n",
    "        a['dev_mean_sq'] = a['dev_mean']**2\n",
    "        a['variance'] = a['dev_mean_sq'].rolling(str(days)+'d').sum()/(a['dev_mean_sq'].rolling(str(days)+'d').count()-1)\n",
    "        # volatility is square root of variance\n",
    "        # do you need annual volatility?\n",
    "        df[name + '_daily_vol'] = np.sqrt(a['variance'])*np.sqrt(252) # 252 trading days in a year\n",
    "        #df.drop(['ln_daily_return'])#,'rol_avg','dev_mean','dev_mean_sq','variance'])\n",
    "        #return df[name + '_daily_vol']\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def stationary_timeseries(df, damned_list = ['volume','std','short','daily']):\n",
    "    # makes columns in df that are not in damned list stationary. Means it calculates the \n",
    "    # difference row by row\n",
    "    columns = df.columns\n",
    "    df_sta = pd.DataFrame()\n",
    "    for column in columns:\n",
    "        if any(elem in column for elem in damned_list):\n",
    "            df_sta[column] = df[column]\n",
    "        else:\n",
    "            df_sta[column+'_sta'] = df[column].shift(1) - df[column]\n",
    "    df_sta.dropna(inplace = True)  # to make sure we have all the data on all the days\n",
    "    return df_sta"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_df(var_name_list,name_list):\n",
    "    # takes list paths of csv files and names and passes it on to get_data.\n",
    "    # Gets df from get_data and merges them on 'Date'\n",
    "    # Calls sigma_fct to get sigma of daily returns \n",
    "    # var_name_list: list of paths to csv files\n",
    "    # name: list of names the columns of df\n",
    "    i=0\n",
    "    for var_name in var_name_list:\n",
    "        if i == 0:\n",
    "            df = get_data(var_name, name_list[i])\n",
    "        else:\n",
    "            df = pd.merge(df,get_data(var_name, name_list[i]),on='Date')\n",
    "        i += 1\n",
    "    df.dropna(inplace = True)  # to make sure we have all the data on all the days\n",
    "    \n",
    "    df = sigma_fct(df,names)\n",
    "    df=df[23:] # drop first entries to get clean results\n",
    "\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_target_data(df, name):\n",
    "    # function to create the target data\n",
    "    # name: name of the column the target data is created of\n",
    "    df[name+'_45d'] = df[name].shift(-32) # stock price in 45 days (approx. 32 trading days)\n",
    "    df.dropna(inplace = True)  # to make sure we have all the data on all the days\n",
    "    # df['short_call'] shows the perfect short call. If you would have shorted a call option with a strike price \n",
    "    # exactly the same as the stock price 45 days later, you made maximum profit. Therefore this is the target \n",
    "    # value for the short call option. The minimum is 0, so we never short a call option with a strike price\n",
    "    # below the stock price right now. The same for the short put option\n",
    "    #df['short_call'] = [x/y-1 if x/y > 1 else 0 for x,y in df[[name+'_45d', name]].values]\n",
    "    #df['short_put'] = [1-x/y if x/y < 1 else 0 for x,y in df[[name+'_45d', name]].values]\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "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>google_open</th>\n",
       "      <th>google_high</th>\n",
       "      <th>google_low</th>\n",
       "      <th>google_close</th>\n",
       "      <th>google</th>\n",
       "      <th>google_volume</th>\n",
       "      <th>google_ra_04</th>\n",
       "      <th>google_std_04</th>\n",
       "      <th>google_ra_09</th>\n",
       "      <th>google_std_09</th>\n",
       "      <th>...</th>\n",
       "      <th>nasdaq_ra_04</th>\n",
       "      <th>nasdaq_std_04</th>\n",
       "      <th>nasdaq_ra_09</th>\n",
       "      <th>nasdaq_std_09</th>\n",
       "      <th>nasdaq_ra_18</th>\n",
       "      <th>nasdaq_std_18</th>\n",
       "      <th>google_daily_vol</th>\n",
       "      <th>s&amp;p_daily_vol</th>\n",
       "      <th>nasdaq_daily_vol</th>\n",
       "      <th>google_45d</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>2008-11-21</th>\n",
       "      <td>130.764862</td>\n",
       "      <td>134.182053</td>\n",
       "      <td>123.188263</td>\n",
       "      <td>130.725006</td>\n",
       "      <td>130.725006</td>\n",
       "      <td>20565700</td>\n",
       "      <td>136.935478</td>\n",
       "      <td>8.750923</td>\n",
       "      <td>143.875715</td>\n",
       "      <td>10.798013</td>\n",
       "      <td>...</td>\n",
       "      <td>1392.540009</td>\n",
       "      <td>68.742234</td>\n",
       "      <td>1452.251430</td>\n",
       "      <td>95.188063</td>\n",
       "      <td>1541.462141</td>\n",
       "      <td>127.530682</td>\n",
       "      <td>0.716470</td>\n",
       "      <td>0.802275</td>\n",
       "      <td>0.738519</td>\n",
       "      <td>156.946732</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2008-11-24</th>\n",
       "      <td>134.127258</td>\n",
       "      <td>134.470963</td>\n",
       "      <td>124.040070</td>\n",
       "      <td>128.239334</td>\n",
       "      <td>128.239334</td>\n",
       "      <td>20184700</td>\n",
       "      <td>129.482170</td>\n",
       "      <td>1.757636</td>\n",
       "      <td>137.580147</td>\n",
       "      <td>9.599201</td>\n",
       "      <td>...</td>\n",
       "      <td>1428.184998</td>\n",
       "      <td>61.992083</td>\n",
       "      <td>1420.705017</td>\n",
       "      <td>68.914436</td>\n",
       "      <td>1498.502503</td>\n",
       "      <td>101.345711</td>\n",
       "      <td>0.529842</td>\n",
       "      <td>0.864400</td>\n",
       "      <td>0.777819</td>\n",
       "      <td>155.761169</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2008-11-25</th>\n",
       "      <td>133.838348</td>\n",
       "      <td>142.794769</td>\n",
       "      <td>133.160873</td>\n",
       "      <td>140.498383</td>\n",
       "      <td>140.498383</td>\n",
       "      <td>21623100</td>\n",
       "      <td>134.368858</td>\n",
       "      <td>8.668457</td>\n",
       "      <td>137.997038</td>\n",
       "      <td>8.831976</td>\n",
       "      <td>...</td>\n",
       "      <td>1468.375000</td>\n",
       "      <td>5.154837</td>\n",
       "      <td>1426.994298</td>\n",
       "      <td>65.073432</td>\n",
       "      <td>1483.279999</td>\n",
       "      <td>90.035297</td>\n",
       "      <td>0.833833</td>\n",
       "      <td>0.791313</td>\n",
       "      <td>0.710049</td>\n",
       "      <td>156.573120</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2008-11-26</th>\n",
       "      <td>139.616684</td>\n",
       "      <td>147.178345</td>\n",
       "      <td>137.584305</td>\n",
       "      <td>145.499634</td>\n",
       "      <td>145.499634</td>\n",
       "      <td>12760800</td>\n",
       "      <td>138.079117</td>\n",
       "      <td>8.880829</td>\n",
       "      <td>137.425609</td>\n",
       "      <td>8.059325</td>\n",
       "      <td>...</td>\n",
       "      <td>1489.616659</td>\n",
       "      <td>36.971751</td>\n",
       "      <td>1434.144287</td>\n",
       "      <td>74.235572</td>\n",
       "      <td>1487.035382</td>\n",
       "      <td>87.259170</td>\n",
       "      <td>0.812863</td>\n",
       "      <td>0.758392</td>\n",
       "      <td>0.707779</td>\n",
       "      <td>149.923050</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2008-11-28</th>\n",
       "      <td>144.747452</td>\n",
       "      <td>147.671494</td>\n",
       "      <td>143.601746</td>\n",
       "      <td>145.933014</td>\n",
       "      <td>145.933014</td>\n",
       "      <td>5150200</td>\n",
       "      <td>143.977010</td>\n",
       "      <td>3.020363</td>\n",
       "      <td>136.698456</td>\n",
       "      <td>8.236770</td>\n",
       "      <td>...</td>\n",
       "      <td>1510.799967</td>\n",
       "      <td>39.935485</td>\n",
       "      <td>1450.814982</td>\n",
       "      <td>86.016593</td>\n",
       "      <td>1480.791532</td>\n",
       "      <td>79.789023</td>\n",
       "      <td>0.871477</td>\n",
       "      <td>0.804582</td>\n",
       "      <td>0.756923</td>\n",
       "      <td>148.936752</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 40 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            google_open  google_high  google_low  google_close      google  \\\n",
       "Date                                                                         \n",
       "2008-11-21   130.764862   134.182053  123.188263    130.725006  130.725006   \n",
       "2008-11-24   134.127258   134.470963  124.040070    128.239334  128.239334   \n",
       "2008-11-25   133.838348   142.794769  133.160873    140.498383  140.498383   \n",
       "2008-11-26   139.616684   147.178345  137.584305    145.499634  145.499634   \n",
       "2008-11-28   144.747452   147.671494  143.601746    145.933014  145.933014   \n",
       "\n",
       "            google_volume  google_ra_04  google_std_04  google_ra_09  \\\n",
       "Date                                                                   \n",
       "2008-11-21       20565700    136.935478       8.750923    143.875715   \n",
       "2008-11-24       20184700    129.482170       1.757636    137.580147   \n",
       "2008-11-25       21623100    134.368858       8.668457    137.997038   \n",
       "2008-11-26       12760800    138.079117       8.880829    137.425609   \n",
       "2008-11-28        5150200    143.977010       3.020363    136.698456   \n",
       "\n",
       "            google_std_09  ...  nasdaq_ra_04  nasdaq_std_04  nasdaq_ra_09  \\\n",
       "Date                       ...                                              \n",
       "2008-11-21      10.798013  ...   1392.540009      68.742234   1452.251430   \n",
       "2008-11-24       9.599201  ...   1428.184998      61.992083   1420.705017   \n",
       "2008-11-25       8.831976  ...   1468.375000       5.154837   1426.994298   \n",
       "2008-11-26       8.059325  ...   1489.616659      36.971751   1434.144287   \n",
       "2008-11-28       8.236770  ...   1510.799967      39.935485   1450.814982   \n",
       "\n",
       "            nasdaq_std_09  nasdaq_ra_18  nasdaq_std_18  google_daily_vol  \\\n",
       "Date                                                                       \n",
       "2008-11-21      95.188063   1541.462141     127.530682          0.716470   \n",
       "2008-11-24      68.914436   1498.502503     101.345711          0.529842   \n",
       "2008-11-25      65.073432   1483.279999      90.035297          0.833833   \n",
       "2008-11-26      74.235572   1487.035382      87.259170          0.812863   \n",
       "2008-11-28      86.016593   1480.791532      79.789023          0.871477   \n",
       "\n",
       "            s&p_daily_vol  nasdaq_daily_vol  google_45d  \n",
       "Date                                                     \n",
       "2008-11-21       0.802275          0.738519  156.946732  \n",
       "2008-11-24       0.864400          0.777819  155.761169  \n",
       "2008-11-25       0.791313          0.710049  156.573120  \n",
       "2008-11-26       0.758392          0.707779  149.923050  \n",
       "2008-11-28       0.804582          0.756923  148.936752  \n",
       "\n",
       "[5 rows x 40 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "var_names = ['GOOG','S&P','Nasdaq2']\n",
    "names = ['google','s&p','nasdaq']\n",
    "df = get_df(var_names, names)\n",
    "df = get_target_data(df, 'google')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "backup['df'] = df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Price of options\n",
    "It is really hard to get data about historic stock options. Be it as api channel or csv download. I managed only to find sources to pay: 500 USD for every year. Because that is too expensive for this project I decided to calculate the option price with the **Black–Scholes formula**, which gives a theoretical estimate of the price of European-style options.\n",
    "\n",
    "Based on the formula for non-divident paying options I calculate the option prices and add a small random term to create a bit of noise."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def black_scholes(S, K, sigma, r=0.03, T=45/365, option = 'call'):\n",
    "    \n",
    "    #S: spot price\n",
    "    #K: strike price\n",
    "    #T: time to maturity in years, 45 days as default\n",
    "    #r: risk-free interest rate, assumed to be constant between t and T\n",
    "    #sigma: standard deviation (volatility) of RETURNS of underlying asset\n",
    "    \n",
    "    d1 = (np.log(S / K) + (r + 0.5 * sigma ** 2) * T) / (sigma * np.sqrt(T))\n",
    "    #d2 = (np.log(S / K) + (r - 0.5 * sigma ** 2) * T) / (sigma * np.sqrt(T))\n",
    "    d2 = d1 - sigma * np.sqrt(T)\n",
    "    \n",
    "    if option == 'call':\n",
    "        result = (S * si.norm.cdf(d1, 0.0, 1.0) - K * np.exp(-r * T) * si.norm.cdf(d2, 0.0, 1.0))\n",
    "    if option == 'put':\n",
    "        result = (K * np.exp(-r * T) * si.norm.cdf(-d2, 0.0, 1.0) - S * si.norm.cdf(-d1, 0.0, 1.0))\n",
    "        \n",
    "    return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def option_prices(date, short_spread, long_spread, strike_days = 45):\n",
    "    # calculates prices of options and calculates initial return (net premium)\n",
    "    date = date_fct(date)\n",
    "    P_s_strike_price, P_l_strike_price, C_s_strike_price, C_l_strike_price = strike_prices(date, short_spread, long_spread)\n",
    "    sigma = get_sigma(date)\n",
    "    stock_price = stock_price_fct(date)\n",
    "    P_s_price = black_scholes(stock_price, P_s_strike_price, sigma, T=strike_days/365, option = 'put')\n",
    "    P_l_price = black_scholes(stock_price, P_l_strike_price, sigma, T=strike_days/365, option = 'put')\n",
    "    C_s_price = black_scholes(stock_price, C_s_strike_price, sigma, T=strike_days/365, option = 'call')\n",
    "    C_l_price = black_scholes(stock_price, C_l_strike_price, sigma, T=strike_days/365, option = 'call')\n",
    "    initial_return = P_s_price + C_s_price - P_l_price - C_l_price\n",
    "    \n",
    "    return P_s_price, P_l_price, C_s_price, C_l_price, initial_return"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now it is time to calculate the strike prices and option prices of all 4 options and the initial return by setting up the iron condor"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Calculate return of options\n",
    "There are five different regimes for the return. R: return, P(C) put(call) option, s(l): short(long), ir: initial return, stock_price: stock price\n",
    "- stock price is below long put option: R = ir + P_l_strike_price - P_s_strike_price\n",
    "- stock price is between long and short put option: R = ir + stock_price - P_s_strike_price\n",
    "- stock price is between short call and short put option: R = ir\n",
    "- stock price is between short and long call option: R = ir - stock_price + C_s_strike price\n",
    "- stock price is above long call option: R = ir - C_l_strike_price + C_s_strike_price"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Define return function of iron condor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def ic_return(trading_date, short_spread, long_spread, strike_days = 45):\n",
    "    P_s_strike_price, P_l_strike_price, C_s_strike_price, C_l_strike_price = strike_prices(trading_date, short_spread, long_spread)\n",
    "    P_s_price, P_l_price, C_s_price, C_l_price, initial_return = option_prices(trading_date,short_spread, long_spread, strike_days)\n",
    "    strike_date = date_fct(trading_date, 45)\n",
    "    final_stock_price = stock_price_fct(strike_date)\n",
    "    if final_stock_price < P_l_strike_price:\n",
    "        final_result = initial_return + P_l_strike_price - P_s_strike_price\n",
    "    elif final_stock_price >= P_l_strike_price and final_stock_price < P_s_strike_price:\n",
    "        final_result = initial_return + stock_price - P_s_strike_price\n",
    "    elif final_stock_price >= P_s_strike_price and final_stock_price < C_s_strike_price:\n",
    "        final_result = initial_return\n",
    "    elif final_stock_price >= C_s_strike_price and final_stock_price < C_l_strike_price:\n",
    "        final_result = initial_return - final_stock_price + C_s_strike_price\n",
    "    elif final_stock_price >= C_l_strike_price:\n",
    "        final_result = initial_return - C_l_strike_price + C_s_strike_price\n",
    "    maximum_result = initial_return\n",
    "    return final_result"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Scale data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/matthias/anaconda3/lib/python3.7/site-packages/sklearn/preprocessing/data.py:334: DataConversionWarning: Data with input dtype int64, float64 were all converted to float64 by MinMaxScaler.\n",
      "  return self.partial_fit(X, y)\n"
     ]
    }
   ],
   "source": [
    "scaler = preprocessing.MinMaxScaler(feature_range=(-1, 1))\n",
    "df_scaled = pd.DataFrame(scaler.fit_transform(df), columns=df.columns, index = df.index)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Split into train and test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_train_test_data(data, train_test_split, batch_size):\n",
    "    # data is dataframe to get train test data out of\n",
    "    train_size_raw = data.shape[0]*train_test_split\n",
    "    train_size = int(train_size_raw - train_size_raw % batch_size) # make train_size multiple of batch_size for \"statefull = True\" in LSTM\n",
    "    train_data = data.iloc[0:train_size]\n",
    "    test_size_raw = data.shape[0] - train_size\n",
    "    test_size = int(test_size_raw - test_size_raw%batch_size) # multiple of batch_size\n",
    "    test_data = data.iloc[train_size : train_size + test_size]\n",
    "    x_train = train_data.drop(columns=['google_45d']) # google_45d is not for training, just for reversing stationarity\n",
    "    y_train = train_data['google_45d']\n",
    "    x_test = test_data.drop(columns=['google_45d'])\n",
    "    y_test = test_data['google_45d']\n",
    "    train_index=x_train.index\n",
    "    x_columns = x_train.columns\n",
    "    test_index = x_test.index\n",
    "    return x_train, x_test, y_train, y_test, test_data, test_index, x_columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 10\n",
    "train_test_split = 0.9\n",
    "x_train, x_test, y_train, y_test, test_data, test_index, x_columns = get_train_test_data(df_scaled, train_test_split, batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Sequential model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/matthias/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:74: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
      "\n",
      "WARNING:tensorflow:From /home/matthias/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
      "\n",
      "WARNING:tensorflow:From /home/matthias/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:4138: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
      "\n",
      "WARNING:tensorflow:From /home/matthias/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:133: The name tf.placeholder_with_default is deprecated. Please use tf.compat.v1.placeholder_with_default instead.\n",
      "\n",
      "WARNING:tensorflow:From /home/matthias/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
      "WARNING:tensorflow:From /home/matthias/anaconda3/lib/python3.7/site-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
      "\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, 128)               5120      \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 16)                2064      \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 4)                 68        \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 4)                 0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 1)                 5         \n",
      "=================================================================\n",
      "Total params: 7,257\n",
      "Trainable params: 7,257\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Build the model architecture\n",
    "seq_model = Sequential()\n",
    "seq_model.add(Dense(128, activation='relu', input_shape=(x_train.shape[1],)))\n",
    "seq_model.add(Dropout(0.2))\n",
    "seq_model.add(Dense(16, activation='relu'))\n",
    "seq_model.add(Dropout(0.2))\n",
    "seq_model.add(Dense(4, activation='relu'))\n",
    "seq_model.add(Dropout(0.2))\n",
    "seq_model.add(Dense(1))\n",
    "\n",
    "# Compile the model using a loss function and an optimizer.\n",
    "seq_model.compile(optimizer='adam',loss='mean_squared_error')\n",
    "seq_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rolling_seq_fit_pred(model,x_test,y_test,x_train,y_train):\n",
    "    predictions = []\n",
    "    epochs = 200\n",
    "    model.fit(x_train,y_train, epochs=epochs, batch_size=50, verbose=0);\n",
    "    counter = 0\n",
    "    for i in x_test.index:\n",
    "        counter+=1\n",
    "        if counter%5==0:  # every 5 trading days the model is retrained\n",
    "            model.reset_states()\n",
    "            model.fit(x_train,y_train, epochs=epochs, batch_size=50, verbose=0);\n",
    "        # make one-step forecast\n",
    "        X = x_test.loc[i].to_numpy().reshape(1,x_test.shape[1]) # reshape single line to have just 1 row\n",
    "        y = y_test.loc[i]\n",
    "        yhat = model.predict(X).reshape(1) # reshape prediction to be just a 1D array\n",
    "        predictions.append(yhat[0])\n",
    "        # add current x_test and y_test line to training data\n",
    "        if i != x_test.index[-1]:\n",
    "            x_train.loc[i] = x_test.loc[i]\n",
    "            y_train.loc[i] = y_test.loc[i]\n",
    "    return predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/matthias/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:986: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/matthias/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:19: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "841.8499984741211"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "start_seq = time.time()\n",
    "yhat = rolling_seq_fit_pred(seq_model,x_test,y_test,x_train,y_train)\n",
    "raw_seq_predictions = pd.DataFrame(yhat,columns = ['seq_prediction'], index = test_index)\n",
    "end_seq = time.time()\n",
    "time_seq = end_seq - start_seq\n",
    "time_seq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Invert scaling for prediction data\n",
    "x_test = pd.DataFrame(x_test, columns = x_columns, index = test_index)\n",
    "unscaled_seq_predictions = pd.concat([x_test, raw_seq_predictions], axis=1)\n",
    "#unscaled_seq_predictions = pd.concat([unscaled_seq_predictions, test_data.google_45d], axis=1)\n",
    "unscaled_seq_predictions = pd.DataFrame(scaler.inverse_transform(unscaled_seq_predictions), columns=unscaled_seq_predictions.columns, index=unscaled_seq_predictions.index)\n",
    "\n",
    "# Invert scaling for actual data\n",
    "unscaled_data = pd.concat([x_test, y_test], axis=1)\n",
    "#unscaled_data = pd.concat([unscaled_data, test_data.google_45d], axis=1)\n",
    "unscaled_data = pd.DataFrame(scaler.inverse_transform(unscaled_data), columns=unscaled_data.columns, index=unscaled_data.index)\n",
    "#unscaled_data = unscaled_data['google_45d_sta']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "backup['unscaled_seq_predictions'] = unscaled_seq_predictions\n",
    "backup['unscaled_data'] = unscaled_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "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>google_open</th>\n",
       "      <th>google_high</th>\n",
       "      <th>google_low</th>\n",
       "      <th>google_close</th>\n",
       "      <th>google</th>\n",
       "      <th>google_volume</th>\n",
       "      <th>google_ra_04</th>\n",
       "      <th>google_std_04</th>\n",
       "      <th>google_ra_09</th>\n",
       "      <th>google_std_09</th>\n",
       "      <th>...</th>\n",
       "      <th>nasdaq_ra_04</th>\n",
       "      <th>nasdaq_std_04</th>\n",
       "      <th>nasdaq_ra_09</th>\n",
       "      <th>nasdaq_std_09</th>\n",
       "      <th>nasdaq_ra_18</th>\n",
       "      <th>nasdaq_std_18</th>\n",
       "      <th>google_daily_vol</th>\n",
       "      <th>s&amp;p_daily_vol</th>\n",
       "      <th>nasdaq_daily_vol</th>\n",
       "      <th>seq_prediction</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>2018-07-31</th>\n",
       "      <td>1220.010010</td>\n",
       "      <td>1227.588013</td>\n",
       "      <td>1205.599976</td>\n",
       "      <td>1217.260010</td>\n",
       "      <td>1217.260010</td>\n",
       "      <td>1644700.0</td>\n",
       "      <td>1218.500000</td>\n",
       "      <td>1.753611</td>\n",
       "      <td>1237.301409</td>\n",
       "      <td>24.147664</td>\n",
       "      <td>...</td>\n",
       "      <td>7650.895020</td>\n",
       "      <td>29.550020</td>\n",
       "      <td>7786.610073</td>\n",
       "      <td>109.258547</td>\n",
       "      <td>7805.587565</td>\n",
       "      <td>85.060270</td>\n",
       "      <td>0.307801</td>\n",
       "      <td>0.089192</td>\n",
       "      <td>0.152314</td>\n",
       "      <td>1054.129204</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-01</th>\n",
       "      <td>1228.000000</td>\n",
       "      <td>1233.469971</td>\n",
       "      <td>1210.209961</td>\n",
       "      <td>1220.010010</td>\n",
       "      <td>1220.010010</td>\n",
       "      <td>1567200.0</td>\n",
       "      <td>1219.003337</td>\n",
       "      <td>1.515790</td>\n",
       "      <td>1239.374268</td>\n",
       "      <td>21.433011</td>\n",
       "      <td>...</td>\n",
       "      <td>7669.693359</td>\n",
       "      <td>38.687654</td>\n",
       "      <td>7767.384347</td>\n",
       "      <td>109.753646</td>\n",
       "      <td>7798.026217</td>\n",
       "      <td>85.881161</td>\n",
       "      <td>0.285090</td>\n",
       "      <td>0.083101</td>\n",
       "      <td>0.146205</td>\n",
       "      <td>1058.003240</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-02</th>\n",
       "      <td>1205.900024</td>\n",
       "      <td>1229.880005</td>\n",
       "      <td>1204.790039</td>\n",
       "      <td>1226.150024</td>\n",
       "      <td>1226.150024</td>\n",
       "      <td>1531300.0</td>\n",
       "      <td>1220.790009</td>\n",
       "      <td>3.781604</td>\n",
       "      <td>1236.241420</td>\n",
       "      <td>21.550840</td>\n",
       "      <td>...</td>\n",
       "      <td>7702.942505</td>\n",
       "      <td>73.619602</td>\n",
       "      <td>7761.944336</td>\n",
       "      <td>106.402575</td>\n",
       "      <td>7798.359340</td>\n",
       "      <td>82.521365</td>\n",
       "      <td>0.271257</td>\n",
       "      <td>0.086055</td>\n",
       "      <td>0.164864</td>\n",
       "      <td>1071.019059</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-03</th>\n",
       "      <td>1229.619995</td>\n",
       "      <td>1230.000000</td>\n",
       "      <td>1215.060059</td>\n",
       "      <td>1223.709961</td>\n",
       "      <td>1223.709961</td>\n",
       "      <td>1089600.0</td>\n",
       "      <td>1221.782501</td>\n",
       "      <td>3.932133</td>\n",
       "      <td>1230.528564</td>\n",
       "      <td>18.079647</td>\n",
       "      <td>...</td>\n",
       "      <td>7748.444946</td>\n",
       "      <td>69.648419</td>\n",
       "      <td>7744.768555</td>\n",
       "      <td>81.003961</td>\n",
       "      <td>7798.808594</td>\n",
       "      <td>82.581619</td>\n",
       "      <td>0.214106</td>\n",
       "      <td>0.085981</td>\n",
       "      <td>0.165144</td>\n",
       "      <td>1071.552692</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-06</th>\n",
       "      <td>1225.000000</td>\n",
       "      <td>1226.088013</td>\n",
       "      <td>1215.796997</td>\n",
       "      <td>1224.770020</td>\n",
       "      <td>1224.770020</td>\n",
       "      <td>1081700.0</td>\n",
       "      <td>1224.239991</td>\n",
       "      <td>0.749575</td>\n",
       "      <td>1221.940003</td>\n",
       "      <td>3.444802</td>\n",
       "      <td>...</td>\n",
       "      <td>7835.844971</td>\n",
       "      <td>33.708070</td>\n",
       "      <td>7747.243327</td>\n",
       "      <td>90.494779</td>\n",
       "      <td>7792.345052</td>\n",
       "      <td>87.849464</td>\n",
       "      <td>0.151643</td>\n",
       "      <td>0.065983</td>\n",
       "      <td>0.139857</td>\n",
       "      <td>1050.835252</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 40 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            google_open  google_high   google_low  google_close       google  \\\n",
       "Date                                                                           \n",
       "2018-07-31  1220.010010  1227.588013  1205.599976   1217.260010  1217.260010   \n",
       "2018-08-01  1228.000000  1233.469971  1210.209961   1220.010010  1220.010010   \n",
       "2018-08-02  1205.900024  1229.880005  1204.790039   1226.150024  1226.150024   \n",
       "2018-08-03  1229.619995  1230.000000  1215.060059   1223.709961  1223.709961   \n",
       "2018-08-06  1225.000000  1226.088013  1215.796997   1224.770020  1224.770020   \n",
       "\n",
       "            google_volume  google_ra_04  google_std_04  google_ra_09  \\\n",
       "Date                                                                   \n",
       "2018-07-31      1644700.0   1218.500000       1.753611   1237.301409   \n",
       "2018-08-01      1567200.0   1219.003337       1.515790   1239.374268   \n",
       "2018-08-02      1531300.0   1220.790009       3.781604   1236.241420   \n",
       "2018-08-03      1089600.0   1221.782501       3.932133   1230.528564   \n",
       "2018-08-06      1081700.0   1224.239991       0.749575   1221.940003   \n",
       "\n",
       "            google_std_09  ...  nasdaq_ra_04  nasdaq_std_04  nasdaq_ra_09  \\\n",
       "Date                       ...                                              \n",
       "2018-07-31      24.147664  ...   7650.895020      29.550020   7786.610073   \n",
       "2018-08-01      21.433011  ...   7669.693359      38.687654   7767.384347   \n",
       "2018-08-02      21.550840  ...   7702.942505      73.619602   7761.944336   \n",
       "2018-08-03      18.079647  ...   7748.444946      69.648419   7744.768555   \n",
       "2018-08-06       3.444802  ...   7835.844971      33.708070   7747.243327   \n",
       "\n",
       "            nasdaq_std_09  nasdaq_ra_18  nasdaq_std_18  google_daily_vol  \\\n",
       "Date                                                                       \n",
       "2018-07-31     109.258547   7805.587565      85.060270          0.307801   \n",
       "2018-08-01     109.753646   7798.026217      85.881161          0.285090   \n",
       "2018-08-02     106.402575   7798.359340      82.521365          0.271257   \n",
       "2018-08-03      81.003961   7798.808594      82.581619          0.214106   \n",
       "2018-08-06      90.494779   7792.345052      87.849464          0.151643   \n",
       "\n",
       "            s&p_daily_vol  nasdaq_daily_vol  seq_prediction  \n",
       "Date                                                         \n",
       "2018-07-31       0.089192          0.152314     1054.129204  \n",
       "2018-08-01       0.083101          0.146205     1058.003240  \n",
       "2018-08-02       0.086055          0.164864     1071.019059  \n",
       "2018-08-03       0.085981          0.165144     1071.552692  \n",
       "2018-08-06       0.065983          0.139857     1050.835252  \n",
       "\n",
       "[5 rows x 40 columns]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unscaled_seq_predictions.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "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>google_open</th>\n",
       "      <th>google_high</th>\n",
       "      <th>google_low</th>\n",
       "      <th>google_close</th>\n",
       "      <th>google</th>\n",
       "      <th>google_volume</th>\n",
       "      <th>google_ra_04</th>\n",
       "      <th>google_std_04</th>\n",
       "      <th>google_ra_09</th>\n",
       "      <th>google_std_09</th>\n",
       "      <th>...</th>\n",
       "      <th>nasdaq_ra_04</th>\n",
       "      <th>nasdaq_std_04</th>\n",
       "      <th>nasdaq_ra_09</th>\n",
       "      <th>nasdaq_std_09</th>\n",
       "      <th>nasdaq_ra_18</th>\n",
       "      <th>nasdaq_std_18</th>\n",
       "      <th>google_daily_vol</th>\n",
       "      <th>s&amp;p_daily_vol</th>\n",
       "      <th>nasdaq_daily_vol</th>\n",
       "      <th>google_45d</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>2018-07-31</th>\n",
       "      <td>1220.010010</td>\n",
       "      <td>1227.588013</td>\n",
       "      <td>1205.599976</td>\n",
       "      <td>1217.260010</td>\n",
       "      <td>1217.260010</td>\n",
       "      <td>1644700.0</td>\n",
       "      <td>1218.500000</td>\n",
       "      <td>1.753611</td>\n",
       "      <td>1237.301409</td>\n",
       "      <td>24.147664</td>\n",
       "      <td>...</td>\n",
       "      <td>7650.895020</td>\n",
       "      <td>29.550020</td>\n",
       "      <td>7786.610073</td>\n",
       "      <td>109.258547</td>\n",
       "      <td>7805.587565</td>\n",
       "      <td>85.060270</td>\n",
       "      <td>0.307801</td>\n",
       "      <td>0.089192</td>\n",
       "      <td>0.152314</td>\n",
       "      <td>1156.050049</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-01</th>\n",
       "      <td>1228.000000</td>\n",
       "      <td>1233.469971</td>\n",
       "      <td>1210.209961</td>\n",
       "      <td>1220.010010</td>\n",
       "      <td>1220.010010</td>\n",
       "      <td>1567200.0</td>\n",
       "      <td>1219.003337</td>\n",
       "      <td>1.515790</td>\n",
       "      <td>1239.374268</td>\n",
       "      <td>21.433011</td>\n",
       "      <td>...</td>\n",
       "      <td>7669.693359</td>\n",
       "      <td>38.687654</td>\n",
       "      <td>7767.384347</td>\n",
       "      <td>109.753646</td>\n",
       "      <td>7798.026217</td>\n",
       "      <td>85.881161</td>\n",
       "      <td>0.285090</td>\n",
       "      <td>0.083101</td>\n",
       "      <td>0.146205</td>\n",
       "      <td>1161.219971</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-02</th>\n",
       "      <td>1205.900024</td>\n",
       "      <td>1229.880005</td>\n",
       "      <td>1204.790039</td>\n",
       "      <td>1226.150024</td>\n",
       "      <td>1226.150024</td>\n",
       "      <td>1531300.0</td>\n",
       "      <td>1220.790009</td>\n",
       "      <td>3.781604</td>\n",
       "      <td>1236.241420</td>\n",
       "      <td>21.550840</td>\n",
       "      <td>...</td>\n",
       "      <td>7702.942505</td>\n",
       "      <td>73.619602</td>\n",
       "      <td>7761.944336</td>\n",
       "      <td>106.402575</td>\n",
       "      <td>7798.359340</td>\n",
       "      <td>82.521365</td>\n",
       "      <td>0.271257</td>\n",
       "      <td>0.086055</td>\n",
       "      <td>0.164864</td>\n",
       "      <td>1171.089966</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-03</th>\n",
       "      <td>1229.619995</td>\n",
       "      <td>1230.000000</td>\n",
       "      <td>1215.060059</td>\n",
       "      <td>1223.709961</td>\n",
       "      <td>1223.709961</td>\n",
       "      <td>1089600.0</td>\n",
       "      <td>1221.782501</td>\n",
       "      <td>3.932133</td>\n",
       "      <td>1230.528564</td>\n",
       "      <td>18.079647</td>\n",
       "      <td>...</td>\n",
       "      <td>7748.444946</td>\n",
       "      <td>69.648419</td>\n",
       "      <td>7744.768555</td>\n",
       "      <td>81.003961</td>\n",
       "      <td>7798.808594</td>\n",
       "      <td>82.581619</td>\n",
       "      <td>0.214106</td>\n",
       "      <td>0.085981</td>\n",
       "      <td>0.165144</td>\n",
       "      <td>1186.869995</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2018-08-06</th>\n",
       "      <td>1225.000000</td>\n",
       "      <td>1226.088013</td>\n",
       "      <td>1215.796997</td>\n",
       "      <td>1224.770020</td>\n",
       "      <td>1224.770020</td>\n",
       "      <td>1081700.0</td>\n",
       "      <td>1224.239991</td>\n",
       "      <td>0.749575</td>\n",
       "      <td>1221.940003</td>\n",
       "      <td>3.444802</td>\n",
       "      <td>...</td>\n",
       "      <td>7835.844971</td>\n",
       "      <td>33.708070</td>\n",
       "      <td>7747.243327</td>\n",
       "      <td>90.494779</td>\n",
       "      <td>7792.345052</td>\n",
       "      <td>87.849464</td>\n",
       "      <td>0.151643</td>\n",
       "      <td>0.065983</td>\n",
       "      <td>0.139857</td>\n",
       "      <td>1166.089966</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 40 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            google_open  google_high   google_low  google_close       google  \\\n",
       "Date                                                                           \n",
       "2018-07-31  1220.010010  1227.588013  1205.599976   1217.260010  1217.260010   \n",
       "2018-08-01  1228.000000  1233.469971  1210.209961   1220.010010  1220.010010   \n",
       "2018-08-02  1205.900024  1229.880005  1204.790039   1226.150024  1226.150024   \n",
       "2018-08-03  1229.619995  1230.000000  1215.060059   1223.709961  1223.709961   \n",
       "2018-08-06  1225.000000  1226.088013  1215.796997   1224.770020  1224.770020   \n",
       "\n",
       "            google_volume  google_ra_04  google_std_04  google_ra_09  \\\n",
       "Date                                                                   \n",
       "2018-07-31      1644700.0   1218.500000       1.753611   1237.301409   \n",
       "2018-08-01      1567200.0   1219.003337       1.515790   1239.374268   \n",
       "2018-08-02      1531300.0   1220.790009       3.781604   1236.241420   \n",
       "2018-08-03      1089600.0   1221.782501       3.932133   1230.528564   \n",
       "2018-08-06      1081700.0   1224.239991       0.749575   1221.940003   \n",
       "\n",
       "            google_std_09  ...  nasdaq_ra_04  nasdaq_std_04  nasdaq_ra_09  \\\n",
       "Date                       ...                                              \n",
       "2018-07-31      24.147664  ...   7650.895020      29.550020   7786.610073   \n",
       "2018-08-01      21.433011  ...   7669.693359      38.687654   7767.384347   \n",
       "2018-08-02      21.550840  ...   7702.942505      73.619602   7761.944336   \n",
       "2018-08-03      18.079647  ...   7748.444946      69.648419   7744.768555   \n",
       "2018-08-06       3.444802  ...   7835.844971      33.708070   7747.243327   \n",
       "\n",
       "            nasdaq_std_09  nasdaq_ra_18  nasdaq_std_18  google_daily_vol  \\\n",
       "Date                                                                       \n",
       "2018-07-31     109.258547   7805.587565      85.060270          0.307801   \n",
       "2018-08-01     109.753646   7798.026217      85.881161          0.285090   \n",
       "2018-08-02     106.402575   7798.359340      82.521365          0.271257   \n",
       "2018-08-03      81.003961   7798.808594      82.581619          0.214106   \n",
       "2018-08-06      90.494779   7792.345052      87.849464          0.151643   \n",
       "\n",
       "            s&p_daily_vol  nasdaq_daily_vol   google_45d  \n",
       "Date                                                      \n",
       "2018-07-31       0.089192          0.152314  1156.050049  \n",
       "2018-08-01       0.083101          0.146205  1161.219971  \n",
       "2018-08-02       0.086055          0.164864  1171.089966  \n",
       "2018-08-03       0.085981          0.165144  1186.869995  \n",
       "2018-08-06       0.065983          0.139857  1166.089966  \n",
       "\n",
       "[5 rows x 40 columns]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unscaled_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "161.82610947637454"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "seq_accuracy=np.sqrt(metrics.mean_squared_error(unscaled_data.google_45d, unscaled_seq_predictions.seq_prediction))\n",
    "seq_accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/matthias/anaconda3/lib/python3.7/site-packages/pandas/plotting/_converter.py:129: FutureWarning: Using an implicitly registered datetime converter for a matplotlib plotting method. The converter was registered by pandas on import. Future versions of pandas will require you to explicitly register matplotlib converters.\n",
      "\n",
      "To register the converters:\n",
      "\t>>> from pandas.plotting import register_matplotlib_converters\n",
      "\t>>> register_matplotlib_converters()\n",
      "  warnings.warn(msg, FutureWarning)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot prediction vs actual\n",
    "plt.plot(unscaled_data['google_45d'], label='Adjusted Close')\n",
    "plt.plot(unscaled_seq_predictions['seq_prediction'], label='FF predicted')\n",
    "plt.grid()\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### LSTM model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Split into train and test set\n",
    "And here you have to be careful according to https://machinelearningmastery.com/time-series-forecasting-long-short-term-memory-network-python/ :\n",
    "\n",
    "\"A benefit of this type of network is that it can learn and remember over long sequences and does not rely on a pre-specified window lagged observation as input. In Keras, this is referred to as stateful, and involves setting the “stateful” argument to “True” when defining an LSTM layer. By default, an LSTM layer in Keras maintains state between data within one batch. A batch of data is a fixed-sized number of rows from the training dataset that defines how many patterns to process before updating the weights of the network. State in the LSTM layer between batches is cleared by default, therefore we must make the LSTM stateful. This gives us fine-grained control over when state of the LSTM layer is cleared, by calling the reset_states() function.\"\n",
    "\n",
    "But in order to can use \"stateful = True\" the data has to be a multiple of the batch size. Therefore I have to use modulo to get the test size correctly"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 10\n",
    "train_test_split = 0.9\n",
    "x_train, x_test, y_train, y_test, test_data, test_index, x_columns = get_train_test_data(df_scaled, train_test_split, batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train=x_train.to_numpy()\n",
    "y_train=y_train.to_numpy()\n",
    "x_test = x_test.to_numpy()\n",
    "y_test = y_test.to_numpy()\n",
    "# Reshape data for (Sample, Timesteps, Features)\n",
    "x_train_reshaped = x_train.reshape(x_train.shape[0], 1, x_train.shape[1])\n",
    "x_test_reshaped = x_test.reshape(x_test.shape[0], 1, x_test.shape[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_1 (LSTM)                (None, 1, 100)            56000     \n",
      "_________________________________________________________________\n",
      "lstm_2 (LSTM)                (None, 100)               80400     \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 1)                 101       \n",
      "=================================================================\n",
      "Total params: 136,501\n",
      "Trainable params: 136,501\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "lstm_model = Sequential()\n",
    "lstm_model.add(LSTM(units=100, input_shape = (x_train_reshaped.shape[1], x_train_reshaped.shape[2]), dropout = 0.2, return_sequences = True))\n",
    "lstm_model.add(LSTM(units = 100, dropout = 0.2))\n",
    "lstm_model.add(Dense(1))\n",
    "lstm_model.compile(optimizer = 'adam', loss='mse')\n",
    "lstm_model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rolling_lstm_fit_pred(model,x_test,y_test,x_train,y_train):\n",
    "    predictions = []\n",
    "    counter = 0\n",
    "    epochs = 15\n",
    "    x_train_reshaped = x_train.reshape(x_train.shape[0], 1, x_train.shape[1])\n",
    "    lstm_model.fit(x_train_reshaped, y_train, epochs=epochs, batch_size=batch_size, verbose=0);\n",
    "    for i in range(len(x_test)):\n",
    "        counter +=1\n",
    "        if counter%5==0:\n",
    "            # reshape x_train\n",
    "            model.reset_states()\n",
    "            x_train_reshaped = x_train.reshape(x_train.shape[0], 1, x_train.shape[1])\n",
    "            lstm_model.fit(x_train_reshaped, y_train, epochs=epochs, batch_size=batch_size, verbose=0);\n",
    "        # make one-step forecast\n",
    "        X = x_test[i,:] # select single line\n",
    "        X = X.reshape(1, 1, len(X))  # reshape for LSTM\n",
    "        y = y_test[i]\n",
    "        yhat = lstm_model.predict(X, batch_size=1)[0] # reshape prediction to be just a 1D array\n",
    "        # store forecast\n",
    "        predictions.append(yhat[0])\n",
    "        # add current x_test and y_test line to training data\n",
    "        if i != len(x_test):\n",
    "            x_train = np.vstack([x_train, x_test[i]])\n",
    "            y_train = np.append(y_train, np.array([y_test[1]]), axis=0)\n",
    "    return predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Get predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/matthias/anaconda3/lib/python3.7/site-packages/tensorflow/python/ops/math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "1214.0347380638123"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Generate predictions\n",
    "start_lstm = time.time()\n",
    "yhat = rolling_lstm_fit_pred(lstm_model,x_test,y_test,x_train,y_train);\n",
    "raw_lstm_predictions = pd.DataFrame(yhat,columns = ['lstm_prediction'], index = test_index)\n",
    "end_lstm = time.time()\n",
    "time_lstm = end_lstm - start_lstm\n",
    "time_lstm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Convert data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_test = pd.DataFrame(x_test, columns = x_columns, index = test_index)\n",
    "y_test = pd.DataFrame(y_test, columns = ['google_45d'], index = test_index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Invert scaling for prediction data\n",
    "unscaled_lstm_predictions = pd.concat([x_test, raw_lstm_predictions], axis=1)\n",
    "unscaled_lstm_predictions = pd.DataFrame(scaler.inverse_transform(unscaled_lstm_predictions), columns=unscaled_lstm_predictions.columns, index=unscaled_lstm_predictions.index)\n",
    "\n",
    "# Invert scaling for actual data\n",
    "test_data_unscaled = pd.concat([x_test, y_test], axis=1)\n",
    "test_data_unscaled = pd.DataFrame(scaler.inverse_transform(test_data_unscaled), columns=test_data_unscaled.columns, index=test_data_unscaled.index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "backup['unscaled_lstm_predictions'] = unscaled_lstm_predictions\n",
    "backup['test_data_unscaled'] = test_data_unscaled"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy of feedforward model is 161.83 \n",
      "Accuracy of LSTM model is 66.0\n"
     ]
    }
   ],
   "source": [
    "lstm_accuracy=np.sqrt(metrics.mean_squared_error(unscaled_data.google_45d, unscaled_lstm_predictions.lstm_prediction))\n",
    "print('Accuracy of feedforward model is {} \\nAccuracy of LSTM model is {}'.format(round(seq_accuracy,2), round(lstm_accuracy)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/matthias/anaconda3/lib/python3.7/site-packages/pandas/plotting/_converter.py:129: FutureWarning: Using an implicitly registered datetime converter for a matplotlib plotting method. The converter was registered by pandas on import. Future versions of pandas will require you to explicitly register matplotlib converters.\n",
      "\n",
      "To register the converters:\n",
      "\t>>> from pandas.plotting import register_matplotlib_converters\n",
      "\t>>> register_matplotlib_converters()\n",
      "  warnings.warn(msg, FutureWarning)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot prediction vs actual\n",
    "plt.plot(test_data_unscaled['google_45d'], label='Adjusted Close')\n",
    "plt.plot(unscaled_seq_predictions['seq_prediction'], label='FF predicted')\n",
    "plt.plot(unscaled_lstm_predictions['lstm_prediction'], label='LSTM predicted')\n",
    "plt.xlabel('Date')\n",
    "plt.ylabel('USD')\n",
    "plt.grid()\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Run time for feedforward model is 841.8499984741211 seconds \n",
      "Run time for LSTM model is 1214.0347380638123 seconds\n"
     ]
    }
   ],
   "source": [
    "print('Run time for feedforward model is {} seconds \\nRun time for LSTM model is {} seconds'.format(time_seq,time_lstm))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "backup.close()"
   ]
  },
  {
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}