10-football-train.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:32:57.736008Z",
     "start_time": "2020-06-03T15:32:57.723423Z"
    }
   },
   "outputs": [],
   "source": [
    "%reload_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:33:31.501265Z",
     "start_time": "2020-06-03T15:33:29.068943Z"
    }
   },
   "outputs": [],
   "source": [
    "from fastai2.tabular.all import *\n",
    "from fastai2.imports import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:33:39.092264Z",
     "start_time": "2020-06-03T15:33:39.075972Z"
    }
   },
   "outputs": [],
   "source": [
    "from fastai2.layers import L1LossFlat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:33:58.902845Z",
     "start_time": "2020-06-03T15:33:58.887874Z"
    }
   },
   "outputs": [],
   "source": [
    "from utils import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:36:17.645874Z",
     "start_time": "2020-06-03T15:36:17.632393Z"
    }
   },
   "outputs": [],
   "source": [
    "path=Path('ftbl-data')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:36:29.057639Z",
     "start_time": "2020-06-03T15:36:29.016143Z"
    }
   },
   "outputs": [],
   "source": [
    "whole_df = pd.read_pickle(path/'trans_over_1000_final.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:36:37.731553Z",
     "start_time": "2020-06-03T15:36:37.714043Z"
    }
   },
   "outputs": [],
   "source": [
    "cat_vars_tpl = ('season','trs_year','trs_month','trs_day','trs_till_deadline',\n",
    "            'contract_left_months', 'contract_left_years','age',\n",
    "            'is_midseason','is_loan','is_end_of_loan',\n",
    "            'nat_national_name','plr_position_main',\n",
    "            'plr_other_positions','plr_nationality_name',\n",
    "            'plr_other_nationality_name','plr_place_of_birth_country_name',\n",
    "            'plr_foot','plr_height','plr_player_agent','from_club_name','from_club_is_first_team',\n",
    "            'from_clb_place','from_clb_qualified_to','from_clb_is_champion','from_clb_is_cup_winner',\n",
    "            'from_clb_is_promoted','from_clb_lg_name','from_clb_lg_country','from_clb_lg_group',\n",
    "            'from_coach_name', 'from_sport_dir_name',\n",
    "            'to_club_name','to_club_is_first_team','to_clb_place','to_clb_qualified_to',\n",
    "            'to_clb_is_champion','to_clb_is_cup_winner','to_clb_is_promoted','to_clb_lg_name','to_clb_lg_country',\n",
    "            'to_clb_lg_group','to_coach_name', 'to_sport_dir_name',\n",
    "            'plr_position_0','plr_position_1','plr_position_2',\n",
    "            'stats_leag_name_0','stats_leag_grp_0','stats_leag_name_1','stats_leag_grp_1','stats_leag_name_2',\n",
    "            'stats_leag_grp_2')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:36:43.399814Z",
     "start_time": "2020-06-03T15:36:43.381269Z"
    }
   },
   "outputs": [],
   "source": [
    "cont_vars_tpl = ('nat_months_from_debut','nat_matches_played','nat_goals_scored','from_clb_pts_avg',\n",
    "             'from_clb_goals_diff_avg','to_clb_pts_avg','to_clb_goals_diff_avg','plr_apps_0',\n",
    "             'plr_apps_1','plr_apps_2','stats_made_goals_0','stats_conc_gols_0','stats_cards_0',\n",
    "             'stats_minutes_0','stats_team_points_0','stats_made_goals_1','stats_conc_gols_1',\n",
    "             'stats_cards_1','stats_minutes_1','stats_team_points_1','stats_made_goals_2',\n",
    "             'stats_conc_gols_2','stats_cards_2','stats_minutes_2','stats_team_points_2', 'pop_log1p')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:36:48.959082Z",
     "start_time": "2020-06-03T15:36:48.945153Z"
    }
   },
   "outputs": [],
   "source": [
    "cat_vars = list(cat_vars_tpl)\n",
    "cont_vars = list(cont_vars_tpl)\n",
    "all_vars = cat_vars + cont_vars"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### prepare"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:37:04.535789Z",
     "start_time": "2020-06-03T15:37:04.520494Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['player_name',\n",
       " 'market_value',\n",
       " 'fee',\n",
       " 'is_future_transfer',\n",
       " 'plr_place_of_birth_name']"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list_diff(whole_df.columns, cat_vars, cont_vars)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We throw away: effectively unique values ('player_name','plr_place_of_birth_name'), constants ('is_future_transfer' as we have no future transfers here anymore) or values we don't want to use now (transfermarkt's 'market_value')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:37:27.718194Z",
     "start_time": "2020-06-03T15:37:27.702203Z"
    }
   },
   "outputs": [],
   "source": [
    "dep_var = 'fee'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:37:33.755234Z",
     "start_time": "2020-06-03T15:37:33.716160Z"
    }
   },
   "outputs": [],
   "source": [
    "df = whole_df[all_vars + [dep_var]].copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:37:39.153235Z",
     "start_time": "2020-06-03T15:37:39.085480Z"
    }
   },
   "outputs": [],
   "source": [
    "df[f\"{dep_var}_orig\"] = df[f\"{dep_var}\"].copy()\n",
    "df[dep_var] = np.log(df[dep_var])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:37:45.099024Z",
     "start_time": "2020-06-03T15:37:45.081222Z"
    }
   },
   "outputs": [],
   "source": [
    "np.random.seed(1001)\n",
    "ln = len(df)\n",
    "valid_idx = np.random.choice(ln, int(ln*0.2), replace=False)\n",
    "train_idx = np.array(list_diff(list(range(ln)), valid_idx))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:37:49.287916Z",
     "start_time": "2020-06-03T15:37:49.268498Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((#1952) [3059,8006,4942,3346,9351,2891,5024,1180,6588,439...],\n",
       " (#7809) [0,1,2,3,4,5,7,8,9,11...])"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "L(list(valid_idx)), L(list(train_idx))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:37:54.830273Z",
     "start_time": "2020-06-03T15:37:54.816149Z"
    }
   },
   "outputs": [],
   "source": [
    "procs=[FillMissing, Categorify, Normalize]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:38:02.295142Z",
     "start_time": "2020-06-03T15:37:59.230142Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([ 6.2146, 14.7725], device='cuda:0')"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "min_log_y = np.log(500)\n",
    "max_log_y = np.max(df[dep_var])*1.2\n",
    "y_range = torch.tensor([min_log_y, max_log_y], device=default_device())\n",
    "y_range"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:38:08.189105Z",
     "start_time": "2020-06-03T15:38:08.163592Z"
    }
   },
   "outputs": [],
   "source": [
    "def exp_mmape(pred:Tensor, targ:Tensor):\n",
    "    \"Exp mean absolute percentage error between `pred` and `targ`.\"\n",
    "    pred,targ = flatten_check(pred,targ)\n",
    "    pred, targ = torch.exp(pred), torch.exp(targ)\n",
    "    pct_var = (targ - pred)/targ\n",
    "    return torch.abs(pct_var).mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:38:19.486155Z",
     "start_time": "2020-06-03T15:38:19.472216Z"
    }
   },
   "outputs": [],
   "source": [
    "BS = 128"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:38:23.678631Z",
     "start_time": "2020-06-03T15:38:23.629358Z"
    }
   },
   "outputs": [],
   "source": [
    "layers = [1000,500]\n",
    "layers_drop = [0.07, 0.7]\n",
    "emb_drop = 0.7\n",
    "cycles = 40\n",
    "w_decay = 0.7\n",
    "max_lr = 1e-3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:38:29.189512Z",
     "start_time": "2020-06-03T15:38:28.152027Z"
    }
   },
   "outputs": [],
   "source": [
    "ds = TabularPandas(df=df,\n",
    "                  procs=procs,\n",
    "                  cat_names=cat_vars,\n",
    "                  cont_names=cont_vars,\n",
    "                  y_names=dep_var,\n",
    "                  y_block=RegressionBlock(),\n",
    "                  splits=(list(train_idx), list(valid_idx)),\n",
    "                  inplace=False,\n",
    "                  reduce_memory=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:38:33.193517Z",
     "start_time": "2020-06-03T15:38:33.136150Z"
    }
   },
   "outputs": [],
   "source": [
    "dls = ds.dataloaders(bs=BS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:38:39.715899Z",
     "start_time": "2020-06-03T15:38:39.608195Z"
    }
   },
   "outputs": [],
   "source": [
    "np.random.seed(1001)\n",
    "learn = tabular_learner(dls=dls,\n",
    "                       layers=layers,\n",
    "                       ps=layers_drop,\n",
    "                       y_range=y_range, \n",
    "                       metrics=exp_mmape,\n",
    "                       loss_func=exp_mmape,\n",
    "                       wd=w_decay)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:38:51.538777Z",
     "start_time": "2020-06-03T15:38:48.928446Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "SuggestedLRs(lr_min=0.2290867567062378, lr_steep=0.0006918309954926372)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:40:11.943162Z",
     "start_time": "2020-06-03T15:39:19.400649Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>exp_mmape</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>5.004932</td>\n",
       "      <td>7.428215</td>\n",
       "      <td>7.428215</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1.581562</td>\n",
       "      <td>0.516851</td>\n",
       "      <td>0.516851</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.716264</td>\n",
       "      <td>0.526602</td>\n",
       "      <td>0.526602</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.440615</td>\n",
       "      <td>0.508337</td>\n",
       "      <td>0.508337</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.362725</td>\n",
       "      <td>0.506861</td>\n",
       "      <td>0.506861</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.361564</td>\n",
       "      <td>0.487077</td>\n",
       "      <td>0.487077</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.382285</td>\n",
       "      <td>0.534076</td>\n",
       "      <td>0.534076</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.403187</td>\n",
       "      <td>0.514997</td>\n",
       "      <td>0.514997</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.405154</td>\n",
       "      <td>0.504891</td>\n",
       "      <td>0.504891</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.395371</td>\n",
       "      <td>0.483766</td>\n",
       "      <td>0.483766</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.363168</td>\n",
       "      <td>0.501054</td>\n",
       "      <td>0.501054</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.335397</td>\n",
       "      <td>0.487427</td>\n",
       "      <td>0.487427</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.311308</td>\n",
       "      <td>0.492860</td>\n",
       "      <td>0.492860</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.295835</td>\n",
       "      <td>0.450454</td>\n",
       "      <td>0.450454</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.287658</td>\n",
       "      <td>0.473366</td>\n",
       "      <td>0.473366</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.277609</td>\n",
       "      <td>0.498784</td>\n",
       "      <td>0.498784</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.269587</td>\n",
       "      <td>0.468537</td>\n",
       "      <td>0.468537</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.255236</td>\n",
       "      <td>0.466537</td>\n",
       "      <td>0.466537</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.244720</td>\n",
       "      <td>0.492484</td>\n",
       "      <td>0.492484</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.233104</td>\n",
       "      <td>0.504406</td>\n",
       "      <td>0.504406</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.222613</td>\n",
       "      <td>0.510601</td>\n",
       "      <td>0.510601</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.216354</td>\n",
       "      <td>0.482866</td>\n",
       "      <td>0.482866</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.198481</td>\n",
       "      <td>0.501129</td>\n",
       "      <td>0.501129</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.185743</td>\n",
       "      <td>0.518851</td>\n",
       "      <td>0.518851</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.172071</td>\n",
       "      <td>0.475695</td>\n",
       "      <td>0.475695</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.163157</td>\n",
       "      <td>0.485729</td>\n",
       "      <td>0.485729</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.152906</td>\n",
       "      <td>0.470690</td>\n",
       "      <td>0.470690</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.141062</td>\n",
       "      <td>0.493615</td>\n",
       "      <td>0.493615</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.132119</td>\n",
       "      <td>0.474597</td>\n",
       "      <td>0.474597</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.119716</td>\n",
       "      <td>0.478520</td>\n",
       "      <td>0.478520</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.109801</td>\n",
       "      <td>0.500951</td>\n",
       "      <td>0.500951</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.098019</td>\n",
       "      <td>0.493374</td>\n",
       "      <td>0.493374</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>32</td>\n",
       "      <td>0.089969</td>\n",
       "      <td>0.470920</td>\n",
       "      <td>0.470920</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>33</td>\n",
       "      <td>0.082493</td>\n",
       "      <td>0.485729</td>\n",
       "      <td>0.485729</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>34</td>\n",
       "      <td>0.078680</td>\n",
       "      <td>0.481054</td>\n",
       "      <td>0.481054</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>35</td>\n",
       "      <td>0.070862</td>\n",
       "      <td>0.489496</td>\n",
       "      <td>0.489496</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>36</td>\n",
       "      <td>0.062660</td>\n",
       "      <td>0.488235</td>\n",
       "      <td>0.488235</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>37</td>\n",
       "      <td>0.059645</td>\n",
       "      <td>0.489242</td>\n",
       "      <td>0.489242</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>38</td>\n",
       "      <td>0.059661</td>\n",
       "      <td>0.480625</td>\n",
       "      <td>0.480625</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>39</td>\n",
       "      <td>0.058144</td>\n",
       "      <td>0.483491</td>\n",
       "      <td>0.483491</td>\n",
       "      <td>00:01</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(n_epoch=cycles, max_lr=max_lr, wd=w_decay)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:40:43.039162Z",
     "start_time": "2020-06-03T15:40:43.024671Z"
    }
   },
   "outputs": [],
   "source": [
    "export_path = path/\"exports\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:41:01.746399Z",
     "start_time": "2020-06-03T15:41:01.608079Z"
    }
   },
   "outputs": [],
   "source": [
    "learn.export(export_path/\"ftbl_483_mean\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Compare with transfermarkt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:44:11.135687Z",
     "start_time": "2020-06-03T15:44:11.120861Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_error(pred, targ):\n",
    "    acc = exp_mmape(pred=tensor(pred).double(), targ=tensor(targ).double())\n",
    "    return int(float(acc)*100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:42:39.313728Z",
     "start_time": "2020-06-03T15:42:39.297139Z"
    }
   },
   "outputs": [],
   "source": [
    "valid_df = whole_df.iloc[valid_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:43:08.170179Z",
     "start_time": "2020-06-03T15:43:08.152504Z"
    }
   },
   "outputs": [],
   "source": [
    "valid_df_no_loans = valid_df.query('is_loan == False')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:43:17.614983Z",
     "start_time": "2020-06-03T15:43:17.600449Z"
    }
   },
   "outputs": [],
   "source": [
    "actuals_l = np.array(np.log1p(valid_df_no_loans['fee']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:43:25.086898Z",
     "start_time": "2020-06-03T15:43:25.072611Z"
    }
   },
   "outputs": [],
   "source": [
    "tm_preds_l =  np.array(np.log1p(valid_df_no_loans['market_value']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:43:38.602514Z",
     "start_time": "2020-06-03T15:43:38.304667Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "dl = learn.dls.test_dl(valid_df_no_loans)\n",
    "model_preds_l = np.array(learn.get_preds(dl=dl)[0].T[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:43:48.275587Z",
     "start_time": "2020-06-03T15:43:48.254873Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "347"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nan_idx = np.argwhere(np.isnan(tm_preds_l))\n",
    "correct_idx = list_diff(range(0,len(actuals_l)), list(nan_idx.T[0]))\n",
    "len(nan_idx)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:43:53.889054Z",
     "start_time": "2020-06-03T15:43:53.874654Z"
    }
   },
   "outputs": [],
   "source": [
    "actuals_l = actuals_l[correct_idx]\n",
    "tm_preds_l = tm_preds_l[correct_idx]\n",
    "model_preds_l = model_preds_l[correct_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:44:15.526155Z",
     "start_time": "2020-06-03T15:44:15.497084Z"
    }
   },
   "outputs": [],
   "source": [
    "tm_error = get_error(pred=tm_preds_l, targ=actuals_l)\n",
    "model_error = get_error(pred=model_preds_l, targ=actuals_l)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:44:24.820342Z",
     "start_time": "2020-06-03T15:44:24.799321Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'transfermarkt': 47, 'model': 46}"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results = {'transfermarkt':tm_error, 'model':model_error}\n",
    "results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-03T15:44:36.836824Z",
     "start_time": "2020-06-03T15:44:36.744945Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7ff6110c3310>"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "results = pd.DataFrame(results, index=['error'])\n",
    "results.plot.barh()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we see even without loans (cause there are no relevant value from transfermarkt to predict loans) our model is equal (and even slightly better) to transfermart predictions"
   ]
  }
 ],
 "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}