{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Глупый градиентный бенчмарк\n",
    "\n",
    "\n",
    "Берём **только** весовые признаки и скармливаем их бустингу"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# подгружаем все нужные пакеты\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "\n",
    "# для встроенных картинок\n",
    "%pylab inline\n",
    "# чуть покрасивше картинки:\n",
    "pd.set_option('display.mpl_style', 'default')\n",
    "figsize(12, 9)\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "#plt.rcParams['figure.figsize'] = 10, 7.5\n",
    "#plt.rcParams['axes.grid'] = True\n",
    "pd.set_option('display.max_columns', None)\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import matplotlib as mpl\n",
    "mpl.rcParams['font.family'] = 'Ubuntu'\n",
    "\n",
    "plt.rc('text', usetex=False)\n",
    "plt.rc('font', family='serif')\n",
    "plt.rc('font', weight='bold')\n",
    "plt.rc('xtick', labelsize=14) \n",
    "plt.rc('ytick', labelsize=14)\n",
    "\n",
    "# чтобы был русский шрифт\n",
    "from matplotlib import rc\n",
    " \n",
    "font = {'family': 'Droid Sans',\n",
    "        'weight': 'normal'}\n",
    "rc('font', **font)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from scipy.sparse import csr_matrix\n",
    "\n",
    "data = pd.read_csv('train2.csv')\n",
    "\n",
    "ss = csr_matrix((data['sum'].values, (data.id.values - 1, data.date.values - 1)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "((110000,), (110000, 431))\n",
      "((110000,), (110000, 424))\n"
     ]
    }
   ],
   "source": [
    "def split_train_test(ss):\n",
    "    \"\"\"\n",
    "    разделение на целевой вектор и спарс-матрицу\n",
    "    \"\"\"\n",
    "    y, s = ss[:, -7:], ss[:, :-7]\n",
    "    y = np.array(y.todense())\n",
    "    y = (((y > 0).cumsum(axis=1) == 1) * y).sum(axis=1)\n",
    "    print (y.shape, s.shape)\n",
    "    return y, s\n",
    "    \n",
    "y, s = split_train_test(ss)\n",
    "y2, s2 = split_train_test(s)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_features(s):\n",
    "    \"\"\"\n",
    "    генерация признаков по спарс-матрице\n",
    "    \"\"\"\n",
    "    m, n = s.shape\n",
    "    k = n % 7\n",
    "    \n",
    "    f = np.zeros((m, 17*24)) # матрица признаков\n",
    "    \n",
    "    \n",
    "    for i in range(m): # отдельно по каждому пользователю\n",
    "        sh = s[i,:].toarray().ravel()\n",
    "        h = sh[k:].reshape(-1, 7)\n",
    "        g = (((h > 0).cumsum(axis=1) == 1) * h).sum(axis=1) # все первые покупки\n",
    "        \n",
    "        q = h.shape[0]\n",
    "        \n",
    "        def get_weighted(w, q, g):\n",
    "            \"\"\"\n",
    "            для создания признаков весовых-схем\n",
    "            w - веса\n",
    "            q = len(g) - можно убрать\n",
    "            g - перечень сумм\n",
    "            \"\"\"\n",
    "            return (np.dot (w, csr_matrix((np.ones(q), (np.arange(q), g)), shape=(q, 17)).toarray()) / sum(w))\n",
    "        \n",
    "        for deg in [0, 1, 2, 3]: # степени в весовой схеме и... индексы\n",
    "            new_features = get_weighted(np.arange(q) ** deg, q, g)\n",
    "            f[i, (17 * deg):(17 * (deg + 1))] = new_features\n",
    "            f[i, (17*6 + 17 * deg):(17*6 + 17 * (deg + 1))] = new_features / max(new_features)\n",
    "            \n",
    "            \n",
    "        # теперь по всем покупкам вообще\n",
    "        l = len(sh)\n",
    "            \n",
    "        for deg in [0, 1, 2, 3]:\n",
    "            new_features = get_weighted(np.arange(l) ** deg, l, sh)\n",
    "            f[i, (17*12 + 17 * deg):(17*12 + 17 * (deg + 1))] = new_features\n",
    "            f[i, (17*18 + 17 * deg):(17*18 + 17 * (deg + 1))] = new_features / max(new_features)            \n",
    "                        \n",
    "        if mod(i, 10000) == 0:\n",
    "            print (i)\n",
    "        \n",
    "    return (f)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "10000\n",
      "20000\n",
      "30000\n",
      "40000\n",
      "50000\n",
      "60000\n",
      "70000\n",
      "80000\n",
      "90000\n",
      "100000\n",
      "0\n",
      "10000\n",
      "20000\n",
      "30000\n",
      "40000\n",
      "50000\n",
      "60000\n",
      "70000\n",
      "80000\n",
      "90000\n",
      "100000\n"
     ]
    }
   ],
   "source": [
    "X  = get_features(s)\n",
    "X2  = get_features(s2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import lightgbm as lgb\n",
    "from time import time\n",
    "from sklearn.metrics import log_loss\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.model_selection import cross_val_predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "39.3623058796\n"
     ]
    }
   ],
   "source": [
    "e = 0.02 * np.random.randn(*X.shape)\n",
    "\n",
    "tm = time()\n",
    "gbm = lgb.LGBMClassifier(learning_rate=0.1, n_estimators=30, nthread=-1)\n",
    "gbm.fit(X2 + e, y2)\n",
    "a = gbm.predict(X)\n",
    "print (time() - tm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(u'\\u0442\\u043e\\u0447\\u043d\\u043e\\u0441\\u0442\\u044c', 0.39496363636363635)\n"
     ]
    }
   ],
   "source": [
    "print (u'точность', np.mean(a==y))\n",
    "# 0.39483636363636365 0.39496363636363635"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "10000\n",
      "20000\n",
      "30000\n",
      "40000\n",
      "50000\n",
      "60000\n",
      "70000\n",
      "80000\n",
      "90000\n",
      "100000\n"
     ]
    }
   ],
   "source": [
    "X0  = get_features(ss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "gbm = lgb.LGBMClassifier(learning_rate=0.1, n_estimators=30, nthread=-1)\n",
    "gbm.fit(X, y)\n",
    "a = gbm.predict(X0)\n",
    "pd.DataFrame({'id': np.arange(1, 110001), 'sum':a}).to_csv('lgb_benchmark-1.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "gbm = lgb.LGBMClassifier(learning_rate=0.1, n_estimators=30, nthread=-1)\n",
    "XX = np.concatenate([X, X2])\n",
    "# XX = XX + 0.02 * np.random.randn(*XX.shape)\n",
    "gbm.fit(XX, np.concatenate([y, y2]))\n",
    "a = gbm.predict(X0)\n",
    "\n",
    "pd.DataFrame({'id': np.arange(1, 110001), 'sum':a}).to_csv('lgb_benchmark-2.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [default]",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}