H&M Framework for Partitioned Validation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np, pandas as pd, datetime as dt\n",
    "import matplotlib.pyplot as plt; plt.style.use('ggplot')\n",
    "import seaborn as sns\n",
    "from collections import defaultdict\n",
    "\n",
    "def iter_to_str(iterable):\n",
    "    return \" \".join(map(lambda x: str(0) + str(x), iterable))\n",
    "\n",
    "def apk(actual, predicted, k=12):\n",
    "    if len(predicted) > k:\n",
    "        predicted = predicted[:k]\n",
    "    score, nhits = 0.0, 0.0\n",
    "    for i, p in enumerate(predicted):\n",
    "        if p in actual and p not in predicted[:i]:\n",
    "            nhits += 1.0\n",
    "            score += nhits / (i + 1.0)\n",
    "    if not actual:\n",
    "        return 0.0\n",
    "    return score / min(len(actual), k)\n",
    "\n",
    "def mapk(actual, predicted, k=12, return_apks=False):\n",
    "    assert len(actual) == len(predicted)\n",
    "    apks = [apk(ac, pr, k) for ac, pr in zip(actual, predicted) if 0 < len(ac)]\n",
    "    if return_apks:\n",
    "        return apks\n",
    "    return np.mean(apks)\n",
    "\n",
    "def blend(dt, w=[], k=12):\n",
    "    if len(w) == 0:\n",
    "        w = [1] * (len(dt))\n",
    "    preds = []\n",
    "    for i in range(len(w)):\n",
    "        preds.append(dt[i].split())\n",
    "    res = {}\n",
    "    for i in range(len(preds)):\n",
    "        if w[i] < 0:\n",
    "            continue\n",
    "        for n, v in enumerate(preds[i]):\n",
    "            if v in res:\n",
    "                res[v] += (w[i] / (n + 1))\n",
    "            else:\n",
    "                res[v] = (w[i] / (n + 1))    \n",
    "    res = list(dict(sorted(res.items(), key=lambda item: -item[1])).keys())\n",
    "    return ' '.join(res[:k])\n",
    "\n",
    "def prune(pred, ok_set, k=12):\n",
    "    pred = pred.split()\n",
    "    post = []\n",
    "    for item in pred:\n",
    "        if int(item) in ok_set and not item in post:\n",
    "            post.append(item)\n",
    "    return \" \".join(post[:k])\n",
    "\n",
    "def validation(actual, predicted, grouping, score=0, index=-1, ignore=False, figsize=(12, 6)):\n",
    "    # actual, predicted : list of lists\n",
    "    # group : pandas Series\n",
    "    # score : pandas DataFrame\n",
    "    if ignore: return\n",
    "    ap12 = mapk(actual, predicted, return_apks=True)\n",
    "    map12 = round(np.mean(ap12), 6)\n",
    "    if isinstance(score, int): score = pd.DataFrame({g:[] for g in sorted(grouping.unique().tolist())})\n",
    "    if index == -1 : index = score.shape[0]\n",
    "    score.loc[index, \"All\"] = map12\n",
    "    plt.figure(figsize=figsize)\n",
    "    plt.subplot(1, 2, 1); sns.histplot(data=ap12, log_scale=(0, 10), bins=20); plt.title(f\"MAP@12 : {map12}\")\n",
    "    for g in grouping.unique():\n",
    "        map12 = round(mapk(actual[grouping == g], predicted[grouping == g]), 6)\n",
    "        score.loc[index, g] = map12\n",
    "    plt.subplot(1, 2, 2); score[[g for g in grouping.unique()[::-1]] + ['All']].loc[index].plot.barh(); plt.title(f\"MAP@12 of Groups\")\n",
    "    vc = pd.Series(predicted).apply(len).value_counts()\n",
    "    score.loc[index, \"Fill\"] = round(1 - sum(vc[k] * (12 - k) / 12 for k in (set(range(12)) & set(vc.index))) / len(actual), 3) * 100\n",
    "    display(score)\n",
    "    return score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_parquet('transactions_train.parquet')\n",
    "sub = pd.read_csv('../input/h-and-m-personalized-fashion-recommendations/sample_submission.csv')\n",
    "cid = pd.DataFrame(sub.customer_id.apply(lambda s: int(s[-16:], 16)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "group = df.groupby('customer_id').sales_channel_id.mean().round().reset_index()\\\n",
    "    .merge(cid, on='customer_id', how='right').rename(columns={'sales_channel_id':'group'})\n",
    "grouping = group.group.fillna(1.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "val_week = 104\n",
    "# id of week to be used in a validation; set 105 if you would like to create a submission\n",
    "val = df.loc[df.week == val_week].groupby('customer_id').article_id.apply(iter_to_str).reset_index()\\\n",
    "    .merge(cid, on='customer_id', how='right')\n",
    "actual = val.article_id.apply(lambda s: [] if pd.isna(s) else s.split())\n",
    "last_date = df.loc[df.week < val_week].t_dat.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>1.0</th>\n",
       "      <th>2.0</th>\n",
       "      <th>All</th>\n",
       "      <th>Fill</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Last Purchase</th>\n",
       "      <td>0.012807</td>\n",
       "      <td>0.024439</td>\n",
       "      <td>0.020594</td>\n",
       "      <td>28.7</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    1.0       2.0       All  Fill\n",
       "Last Purchase  0.012807  0.024439  0.020594  28.7"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "init_date = last_date - dt.timedelta(days=9999)\n",
    "train = df.loc[(df.t_dat >= init_date) & (df.t_dat <= last_date)].copy()\n",
    "train = train.merge(train.groupby('customer_id').t_dat.max().reset_index().rename(columns={'t_dat':'l_dat'}),\n",
    "                    on = 'customer_id', how='left')\n",
    "train['d_dat'] = (train.l_dat - train.t_dat).dt.days\n",
    "train = train.loc[train.d_dat < 14].sort_values(['t_dat'], ascending=False).drop_duplicates(['customer_id', 'article_id'])\n",
    "sub['last_purchase'] = train.groupby('customer_id')\\\n",
    "    .article_id.apply(iter_to_str).reset_index()\\\n",
    "    .merge(cid, on='customer_id', how='right').article_id.fillna('')\n",
    "\n",
    "predicted = sub['last_purchase'].apply(lambda s: [] if pd.isna(s) else s.split())\n",
    "score = validation(actual, predicted, grouping, index='Last Purchase', ignore=(val_week == 105))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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>1.0</th>\n",
       "      <th>2.0</th>\n",
       "      <th>All</th>\n",
       "      <th>Fill</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Last Purchase</th>\n",
       "      <td>0.012807</td>\n",
       "      <td>0.024439</td>\n",
       "      <td>0.020594</td>\n",
       "      <td>28.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Other Colors</th>\n",
       "      <td>0.005846</td>\n",
       "      <td>0.008675</td>\n",
       "      <td>0.007740</td>\n",
       "      <td>15.4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    1.0       2.0       All  Fill\n",
       "Last Purchase  0.012807  0.024439  0.020594  28.7\n",
       "Other Colors   0.005846  0.008675  0.007740  15.4"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "init_date = last_date - dt.timedelta(days=6)\n",
    "train = df.loc[(df.t_dat >= init_date) & (df.t_dat <= last_date)].copy()\\\n",
    "    .groupby(['article_id']).t_dat.count().reset_index()\n",
    "adf = pd.read_parquet('articles.parquet')\n",
    "adf = adf.merge(train, on='article_id', how='left').rename(columns={'t_dat':'ct'})\\\n",
    "    .sort_values('ct', ascending=False).query('ct > 0')\n",
    "\n",
    "map_to_col = defaultdict(list)\n",
    "for aid in adf.article_id.tolist():\n",
    "    map_to_col[aid] = list(filter(lambda x: x != aid, adf[adf.product_code == aid // 1000].article_id.tolist()))[:1]\n",
    "\n",
    "def map_to_variation(s):\n",
    "    f = lambda item: iter_to_str(map_to_col[int(item)])\n",
    "    return ' '.join(map(f, s.split()))\n",
    "sub['other_colors'] = sub['last_purchase'].fillna('').apply(map_to_variation)\n",
    "\n",
    "predicted = sub['other_colors'].apply(lambda s: [] if pd.isna(s) else s.split())\n",
    "score = validation(actual, predicted, grouping, score, index='Other Colors', ignore=(val_week == 105))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>1.0</th>\n",
       "      <th>2.0</th>\n",
       "      <th>All</th>\n",
       "      <th>Fill</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Last Purchase</th>\n",
       "      <td>0.012807</td>\n",
       "      <td>0.024439</td>\n",
       "      <td>0.020594</td>\n",
       "      <td>28.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Other Colors</th>\n",
       "      <td>0.005846</td>\n",
       "      <td>0.008675</td>\n",
       "      <td>0.007740</td>\n",
       "      <td>15.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Popular Items</th>\n",
       "      <td>0.009144</td>\n",
       "      <td>0.008807</td>\n",
       "      <td>0.008918</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    1.0       2.0       All   Fill\n",
       "Last Purchase  0.012807  0.024439  0.020594   28.7\n",
       "Other Colors   0.005846  0.008675  0.007740   15.4\n",
       "Popular Items  0.009144  0.008807  0.008918  100.0"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "init_date = last_date - dt.timedelta(days=5 - 1)\n",
    "group_df = pd.concat([cid, group.group.fillna(1)], axis=1) # grouping can be changed\n",
    "group_df.columns = ['customer_id', 'group']\n",
    "train = df.loc[(df.t_dat >= init_date) & (df.t_dat <= last_date)].copy()\\\n",
    "    .merge(group_df, on='customer_id', how='left')\\\n",
    "    .groupby(['group', 'article_id']).t_dat.count().reset_index()\n",
    "items = defaultdict(str)\n",
    "for g in train.group.unique():\n",
    "    items[g] = iter_to_str(train.loc[train.group == g].sort_values('t_dat', ascending=False).article_id.tolist()[:12])\n",
    "\n",
    "sub['popular_items'] = group_df.group.map(items)\n",
    "\n",
    "predicted = sub['popular_items'].apply(lambda s: [] if pd.isna(s) else s.split())\n",
    "score = validation(actual, predicted, grouping, score, index='Popular Items', ignore=(val_week == 105))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "init_date = last_date - dt.timedelta(days=11)\n",
    "sold_set = set(df.loc[(df.t_dat >= init_date) & (df.t_dat <= last_date)].article_id.tolist())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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>1.0</th>\n",
       "      <th>2.0</th>\n",
       "      <th>All</th>\n",
       "      <th>Fill</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Last Purchase</th>\n",
       "      <td>0.012807</td>\n",
       "      <td>0.024439</td>\n",
       "      <td>0.020594</td>\n",
       "      <td>28.7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Other Colors</th>\n",
       "      <td>0.005846</td>\n",
       "      <td>0.008675</td>\n",
       "      <td>0.007740</td>\n",
       "      <td>15.4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Popular Items</th>\n",
       "      <td>0.009144</td>\n",
       "      <td>0.008807</td>\n",
       "      <td>0.008918</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Prediction</th>\n",
       "      <td>0.018279</td>\n",
       "      <td>0.029840</td>\n",
       "      <td>0.026018</td>\n",
       "      <td>100.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    1.0       2.0       All   Fill\n",
       "Last Purchase  0.012807  0.024439  0.020594   28.7\n",
       "Other Colors   0.005846  0.008675  0.007740   15.4\n",
       "Popular Items  0.009144  0.008807  0.008918  100.0\n",
       "Prediction     0.018279  0.029840  0.026018  100.0"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sub['prediction'] = sub[['last_purchase', 'other_colors', 'popular_items']]\\\n",
    "    .apply(blend, w=[100, 10, 1], axis=1, k=32).apply(prune, ok_set=sold_set)\n",
    "predicted = sub.prediction.apply(lambda s: [] if pd.isna(s) else s.split())\n",
    "score = validation(actual, predicted, grouping, score, index='Prediction', ignore=(val_week == 105))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "sns.barplot(data=score, x='All', y=score.index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "if val_week == 105: sub[['customer_id', 'prediction']].to_csv('submission.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "f436cdb28f556d638768226e88df829b306d6cce33caefeb33828bb6037ec961"
  },
  "kernelspec": {
   "display_name": "Python 3.8.12 ('kaggle')",
   "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": "3.8.12"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}