BiSPM.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import scipy.sparse as sps\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tqdm import tqdm\n",
    "\n",
    "import sys\n",
    "sys.path.append(\"/Users/alessiorussointroito/Documents/GitHub/Structural-Perturbation-Method\")\n",
    "\n",
    "#from SPM_fast import SPM\n",
    "from BiSPM import BiSPM\n",
    "\n",
    "import similaripy as sim"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_svs = 300\n",
    "k = 10\n",
    "p = 0.3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\n",
    "    \"/Users/alessiorussointroito/Downloads/Telegram Desktop/recommender-system-2020-challenge-polimi/data_train.csv\")\n",
    "\n",
    "le = LabelEncoder()\n",
    "df['new_col'] = le.fit_transform(df.col)\n",
    "\n",
    "row_size = len(df.row.unique())\n",
    "col_size = len(le.classes_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(df.row, df.new_col, test_size=0.20, random_state=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "bip_adj = sps.csr_matrix((np.ones(X_train.shape[0]), (X_train, y_train)), shape=(row_size, col_size))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_indices = X_test.unique()\n",
    "test_indices = np.sort(test_indices)\n",
    "test = pd.DataFrame({'row': X_test, 'target': y_test})\n",
    "\n",
    "targets = test.groupby(test.row)['target'].apply(lambda x: list(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "urm = sim.normalization.bm25(bip_adj)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "bspm = BiSPM(urm, target=test_indices, n_sv=n_svs, p=p)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "SVD:  10%|█         | 1/10 [01:20<12:04, 80.52s/it]                  \n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-9-112193b18389>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mrankings\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mbspm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mk_runs\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mk\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m~/Documents/GitHub/Structural-Perturbation-Method/BiSPM.py\u001b[0m in \u001b[0;36mk_runs\u001b[0;34m(self, k, save)\u001b[0m\n\u001b[1;32m     99\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msplit_init\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0misSingle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    100\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 101\u001b[0;31m             \u001b[0mres\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcompute_perturbed_matrix\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mB_r\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdelta_B\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    102\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    103\u001b[0m             \u001b[0;31m# Reset delta_B\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/Documents/GitHub/Structural-Perturbation-Method/BiSPM.py\u001b[0m in \u001b[0;36mcompute_perturbed_matrix\u001b[0;34m(self, B_r, delta_B)\u001b[0m\n\u001b[1;32m     74\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpbar\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mset_description\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"SVD\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     75\u001b[0m         \u001b[0;31m# SVD\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 76\u001b[0;31m         \u001b[0mu\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ms\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvt\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msps\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinalg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msvds\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mB_r\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mk\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mn_sing_values\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     77\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     78\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpbar\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mset_description\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Delta_svs\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/scipy/sparse/linalg/eigen/arpack/arpack.py\u001b[0m in \u001b[0;36msvds\u001b[0;34m(A, k, ncv, tol, which, v0, maxiter, return_singular_vectors, solver)\u001b[0m\n\u001b[1;32m   1865\u001b[0m     \u001b[0;32melif\u001b[0m \u001b[0msolver\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'arpack'\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0msolver\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1866\u001b[0m         eigvals, eigvec = eigsh(XH_X, k=k, tol=tol ** 2, maxiter=maxiter,\n\u001b[0;32m-> 1867\u001b[0;31m                                 ncv=ncv, which=which, v0=v0)\n\u001b[0m\u001b[1;32m   1868\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1869\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/scipy/sparse/linalg/eigen/arpack/arpack.py\u001b[0m in \u001b[0;36meigsh\u001b[0;34m(A, k, M, sigma, which, v0, ncv, maxiter, tol, return_eigenvectors, Minv, OPinv, mode)\u001b[0m\n\u001b[1;32m   1685\u001b[0m     \u001b[0;32mwith\u001b[0m \u001b[0m_ARPACK_LOCK\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1686\u001b[0m         \u001b[0;32mwhile\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mparams\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconverged\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1687\u001b[0;31m             \u001b[0mparams\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0miterate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1688\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1689\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mparams\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mextract\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mreturn_eigenvectors\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python3.7/site-packages/scipy/sparse/linalg/eigen/arpack/arpack.py\u001b[0m in \u001b[0;36miterate\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    537\u001b[0m             self._arpack_solver(self.ido, self.bmat, self.which, self.k,\n\u001b[1;32m    538\u001b[0m                                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtol\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mresid\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mv\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0miparam\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 539\u001b[0;31m                                 self.ipntr, self.workd, self.workl, self.info)\n\u001b[0m\u001b[1;32m    540\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    541\u001b[0m         \u001b[0mxslice\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mslice\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mipntr\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mipntr\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m-\u001b[0m \u001b[0;36m1\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mn\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "rankings = bspm.k_runs(k=k)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Evaluation\n",
    "\n",
    "def precision(is_relevant, relevant_items):\n",
    "    # is_relevant = np.in1d(recommended_items, relevant_items, assume_unique=True)\n",
    "    precision_score = np.sum(is_relevant, dtype=np.float32) / len(is_relevant)\n",
    "    return precision_score\n",
    "\n",
    "\n",
    "def recall(is_relevant, relevant_items):\n",
    "    # is_relevant = np.in1d(recommended_items, relevant_items, assume_unique=True)\n",
    "    recall_score = np.sum(is_relevant, dtype=np.float32) / len(is_relevant)\n",
    "    return recall_score\n",
    "\n",
    "\n",
    "def MAP(is_relevant, relevant_items):\n",
    "    # Cumulative sum: precision at 1, at 2, at 3 ...\n",
    "    p_at_k = is_relevant * np.cumsum(is_relevant, dtype=np.float32) / (1 + np.arange(is_relevant.shape[0]))\n",
    "    map_score = np.sum(p_at_k) / np.min([len(relevant_items), len(is_relevant)])\n",
    "    return map_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5644/5644 [00:14<00:00, 391.41it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SPM n_eigen = 300 , k = 10  \n",
      " Precision = 0.026009922041105162 \n",
      " Recall = 0.026009922041105162 \n",
      " MAP = 0.04137939172768613\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "n_users = len(test_indices)\n",
    "\n",
    "cumulative_precision = 0.0\n",
    "cumulative_recall = 0.0\n",
    "cumulative_MAP = 0.0\n",
    "num_eval = 0\n",
    "\n",
    "at = 10\n",
    "\n",
    "for i, user_id in enumerate(tqdm(test_indices)):\n",
    "    relevant_items = targets[user_id]\n",
    "    #recommended_items = rankings[user_id].argsort()[::-1][:at]\n",
    "    recommended_items = rankings[i]\n",
    "\n",
    "    # Filter Seen:\n",
    "    # 1. Remove items already seen by the user\n",
    "    seen_indices = sps.find(bip_adj[user_id])[1]    # Con [1] Prendiamo solo le colonne d'interesse della matrice di adiacenza\n",
    "    mask = np.zeros(bip_adj.shape[1], dtype=bool)\n",
    "    mask[seen_indices] = True\n",
    "    \n",
    "    recommended_items[mask] = -np.inf\n",
    "\n",
    "    # Recommend\n",
    "    recommended_items = recommended_items.argsort()[::-1][:at]\n",
    "\n",
    "    num_eval += 1\n",
    "\n",
    "    is_relevant = np.in1d(recommended_items, relevant_items, assume_unique=True)\n",
    "\n",
    "    cumulative_precision += precision(is_relevant, relevant_items)\n",
    "    cumulative_recall += recall(is_relevant, relevant_items)\n",
    "    cumulative_MAP += MAP(is_relevant, relevant_items)\n",
    "\n",
    "cumulative_precision /= num_eval\n",
    "cumulative_recall /= num_eval\n",
    "cumulative_MAP /= num_eval\n",
    "\n",
    "print(f\"SPM n_eigen = {n_svs} , k = {k}  \\n Precision = {cumulative_precision} \\n Recall = {cumulative_recall} \\n MAP = {cumulative_MAP}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "a = np.array([True, True, False, True])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([1, 2, 2, 3])"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.cumsum(a)"
   ]
  },
  {
   "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": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}