Untitled28.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import matplotlib\n",
    "matplotlib.rcParams['axes.unicode_minus'] = False\n",
    "\n",
    "from matplotlib import font_manager, rc\n",
    "font_name = font_manager.FontProperties(fname=\"c:/Windows/Fonts/malgun.ttf\").get_name()\n",
    "rc('font', family=font_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fig = plt.figure(figsize=(100, 6))\n",
    "\n",
    "top_axes = plt.subplot2grid((4,4), (0,0), rowspan=3, colspan=4)\n",
    "bottom_axes = plt.subplot2grid((4,4), (3,0), rowspan=1, colspan=4)\n",
    "bottom_axes.get_yaxis().get_major_formatter().set_scientific(False)\n",
    "\n",
    "top_axes.plot(df.index, df['Close'], label='Close')\n",
    "bottom_axes.plot(df.index, df['cci'], color='r')\n",
    "bottom_axes.plot(df.index, np.ones(len(df.index))*100, color='g')\n",
    "bottom_axes.plot(df.index, np.ones(len(df.index))*0, color='y')\n",
    "bottom_axes.plot(df.index, np.ones(len(df.index))*(-100), color='c')\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "with open('raw_data_20181008.pickle', 'rb') as handle:\n",
    "    raw_data = pickle.load(handle)\n",
    "    \n",
    "copy_keys = list(raw_data.keys())\n",
    "for stock in copy_keys:\n",
    "    if len(raw_data[stock].loc[:'2013-01-01']) < 1:\n",
    "        del raw_data[stock]\n",
    "del copy_keys\n",
    "\n",
    "# stock에서 Volume==0인 index, stock에서 제외\n",
    "for key, df in raw_data.items():\n",
    "    raw_data[key] = df.drop(df.index[np.where(df.Volume==0)[0]], axis=0)\n",
    "    del df['Adj Close']\n",
    "    df.Volume = df.Volume / 1e7"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = raw_data['롯데정밀화학']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys([1])\n",
      "dict_keys([1, 2])\n",
      "dict_keys([1, 2, 3])\n",
      "dict_keys([1, 2, 3, 4])\n",
      "dict_keys([1, 2, 3, 4, 5])\n",
      "dict_keys([1, 2, 3, 4, 5, 6])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29])\n",
      "dict_keys([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30])\n"
     ]
    }
   ],
   "source": [
    "data_dict = {}\n",
    "for k in range(1, 91):\n",
    "    raw = pd.DataFrame(index=df.index)\n",
    "    raw['returnsCloseNext{}'.format(k)] = np.log(df['Close'].shift(-1*k) / df['Open'])\n",
    "\n",
    "    # Moving Average\n",
    "    for j in range(1, k+1):\n",
    "        raw['close_ma_{}'.format(j)] = df.Close.rolling(j).mean()\n",
    "\n",
    "    # Return of Change\n",
    "    raw['roc_1'] = df.Close.pct_change(periods=1)\n",
    "    raw['roc_2'] = df.Close.pct_change(periods=2)\n",
    "\n",
    "    # RSI\n",
    "    U = np.where(df.Close.diff(1) > 0, df.Close.diff(1), 0)\n",
    "    D = np.where(df.Close.diff(1) < 0, df.Close.diff(1)*(-1), 0)\n",
    "    for j in range(1, k+1):\n",
    "        m_N = i\n",
    "        AU = pd.DataFrame(U).rolling(window=m_N, min_periods=m_N).mean()\n",
    "        AD = pd.DataFrame(D).rolling(window=m_N, min_periods=m_N).mean()\n",
    "        RSI = AU.div(AD+AU)*100\n",
    "        raw['rsi_{}'.format(j)] = RSI.values\n",
    "\n",
    "        # CV\n",
    "        if j < 5:\n",
    "            pass\n",
    "        else:\n",
    "            raw['his_vol_{}'.format(j)] = df.Close.rolling(j).std() / df.Close.rolling(j).mean()\n",
    "\n",
    "    \n",
    "    for n in [7, 14, 21, 28]:\n",
    "        # William%R\n",
    "        W_R = -100 * ((df.High.rolling(n).max() - df.Close) / \n",
    "           (df.High.rolling(n).max() - df.Low.rolling(n).min()))\n",
    "        raw['w_r'] = W_R\n",
    "\n",
    "        # Stochastic Oscillator\n",
    "        percent_K = 100 * ((df.Close - df.Low.rolling(n).min()) / \n",
    "           (df.High.rolling(n).max() - df.Low.rolling(n).min()))\n",
    "        sto_slowD = percent_K.rolling(3).mean().rolling(3).mean()\n",
    "        raw['sto_slowD'] = sto_slowD\n",
    "        \n",
    "        # Volume Prce Trend\n",
    "        VPT = pd.Series(index=df.index)\n",
    "        VPT.iloc[0] = 0\n",
    "        for n in range(1, len(VPT)):\n",
    "            VPT.iloc[n] = VPT.iloc[n-1] + ((df.Close.diff(1) / df.Close) * df.Volume).iloc[n]\n",
    "        raw['vpt'] = VPT\n",
    "\n",
    "    # Commodity Channel Index\n",
    "    TP = (df.Close + df.Low + df.High) / 3\n",
    "    for window in [10, 20, 30]:\n",
    "        dataX = []\n",
    "        for i in range(len(TP) - window+1):\n",
    "            a = TP.values[i:(i+window)]\n",
    "            dataX.append(a)\n",
    "        dataMean = [i.mean() for i in dataX]\n",
    "        AVEDEV = [abs(x - mean).mean() for x, mean \n",
    "                  in zip(dataX, dataMean)]\n",
    "        for i in range(window-1):\n",
    "            AVEDEV.insert(0,np.nan)\n",
    "        AVEDEV = pd.Series(AVEDEV, index = TP.index)\n",
    "        CCI = (TP - TP.rolling(window).mean()) / (0.015 * AVEDEV)\n",
    "        raw['cci'] = CCI\n",
    "\n",
    "    # Disparity Index\n",
    "    for j in range(5, 31, 5):\n",
    "        DI = 100 * ((df.Close - df.Close.rolling(j).mean()) / \n",
    "                    df.Close.rolling(j).mean())\n",
    "        raw['di{}'.format(j)] = DI\n",
    "    data_dict[k] = raw\n",
    "    print(data_dict.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "for ix, df2 in data_dict.items():\n",
    "    data_dict[ix] = df2.dropna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i,j in data_dict.items():\n",
    "    data_dict[i] = j.loc['2013-01-01':'2018-07-31']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "from xgboost import XGBRegressor\n",
    "xgb_reg = XGBRegressor()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done : 휴켐스 \n",
      "\t0.1247 sec\n",
      "Done : 휴켐스 \n",
      "\t0.1097 sec\n",
      "Done : 휴켐스 \n",
      "\t0.1416 sec\n",
      "Done : 휴켐스 \n",
      "\t0.1337 sec\n",
      "Done : 휴켐스 \n",
      "\t0.1606 sec\n",
      "Done : 휴켐스 \n",
      "\t0.1735 sec\n",
      "Done : 휴켐스 \n",
      "\t0.2075 sec\n",
      "Done : 휴켐스 \n",
      "\t0.2154 sec\n",
      "Done : 휴켐스 \n",
      "\t0.2344 sec\n",
      "Done : 휴켐스 \n",
      "\t0.2443 sec\n",
      "Done : 휴켐스 \n",
      "\t0.2573 sec\n",
      "Done : 휴켐스 \n",
      "\t0.2972 sec\n",
      "Done : 휴켐스 \n",
      "\t0.3341 sec\n",
      "Done : 휴켐스 \n",
      "\t0.3371 sec\n",
      "Done : 휴켐스 \n",
      "\t0.3271 sec\n",
      "Done : 휴켐스 \n",
      "\t0.3490 sec\n",
      "Done : 휴켐스 \n",
      "\t0.3610 sec\n",
      "Done : 휴켐스 \n",
      "\t0.3930 sec\n",
      "Done : 휴켐스 \n",
      "\t0.4009 sec\n",
      "Done : 휴켐스 \n",
      "\t0.4059 sec\n",
      "Done : 휴켐스 \n",
      "\t0.4528 sec\n",
      "Done : 휴켐스 \n",
      "\t0.4408 sec\n",
      "Done : 휴켐스 \n",
      "\t0.5007 sec\n",
      "Done : 휴켐스 \n",
      "\t0.4727 sec\n",
      "Done : 휴켐스 \n",
      "\t0.4986 sec\n",
      "Done : 휴켐스 \n",
      "\t0.5386 sec\n",
      "Done : 휴켐스 \n",
      "\t0.5336 sec\n",
      "Done : 휴켐스 \n",
      "\t0.5316 sec\n",
      "Done : 휴켐스 \n",
      "\t0.5575 sec\n",
      "Done : 휴켐스 \n",
      "\t0.5715 sec\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "from sklearn import metrics\n",
    "\n",
    "result = {}\n",
    "start = time.time()\n",
    "for i in data_dict.keys():\n",
    "    semi_start = time.time()\n",
    "    y = 'returnsCloseNext{}'.format(i)\n",
    "    X = [col for col in data_dict[i].columns if col.find('Next') == -1]\n",
    "    train_data = data_dict[i].iloc[:int(0.8*len(data_dict[i]))][X].values\n",
    "    train_label = data_dict[i].iloc[:int(0.8*len(data_dict[i]))][y].values\n",
    "    test_data = data_dict[i].iloc[int(0.8*len(data_dict[i])):][X].values\n",
    "    test_label = data_dict[i].iloc[int(0.8*len(data_dict[i])):][y]\n",
    "    xgb_reg = XGBRegressor()\n",
    "    xgb_reg.fit(train_data, train_label)\n",
    "    pred = pd.Series(xgb_reg.predict(test_data), index=test_label.index)\n",
    "    rmse = metrics.mean_squared_error(test_label.values, pred.values) ** .5\n",
    "    accuracy = metrics.accuracy_score(np.where(test_label.values>0, 1, 0), \n",
    "                       np.where(pred.values>0, 1, 0))\n",
    "    result[i] = [rmse, accuracy, pred]\n",
    "    semi_elapse = time.time() - semi_start\n",
    "    print('Done :', key, '\\n\\t{:.4f} sec'.format(semi_elapse))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "rmse = [data[0] for key, data in result.items()]\n",
    "acc = [data[1] for key, data in result.items()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1f7ad1ab5f8>]"
      ]
     },
     "execution_count": 95,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1f7acf724e0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(acc)"
   ]
  },
  {
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}