Averaging_Baseline.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "import ImportData\n",
    "from sklearn.metrics import mean_squared_error\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn import preprocessing, metrics\n",
    "import numpy as np\n",
    "import scipy\n",
    "import copy\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = ImportData.importDataAsTable(False, False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>IC50</th>\n",
       "      <th>cell_line</th>\n",
       "      <th>drug_id</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5.457453</td>\n",
       "      <td>ES3</td>\n",
       "      <td>176870</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6.177945</td>\n",
       "      <td>ES5</td>\n",
       "      <td>176870</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3.954488</td>\n",
       "      <td>EW-11</td>\n",
       "      <td>176870</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>5.429563</td>\n",
       "      <td>NCI-H1395</td>\n",
       "      <td>176870</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6.239001</td>\n",
       "      <td>NCI-H1770</td>\n",
       "      <td>176870</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       IC50  cell_line  drug_id\n",
       "0  5.457453        ES3   176870\n",
       "1  6.177945        ES5   176870\n",
       "2  3.954488      EW-11   176870\n",
       "3  5.429563  NCI-H1395   176870\n",
       "4  6.239001  NCI-H1770   176870"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df, test_df = train_test_split(df, test_size=0.2)\n",
    "y_test = np.array(test_df[\"IC50\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "def f1_score(y_test, y_pred):\n",
    "    y_test_binary = copy.deepcopy(y_test)\n",
    "    y_test_binary = preprocessing.binarize(y_test_binary.reshape(1,-1), threshold=5, copy=False)[0]\n",
    "    y_pred_binary = copy.deepcopy(y_pred)\n",
    "    y_pred_binary = preprocessing.binarize(y_pred_binary.reshape(1,-1), threshold=5, copy=False)[0]\n",
    "    return metrics.f1_score(y_test_binary, y_pred_binary)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Based on the cell_line average"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_cell_line_mean = train_df.groupby(['cell_line'])['IC50'].mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_cell_line = np.array([train_cell_line_mean[cell_line] for cell_line in test_df[\"cell_line\"]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3.1944096507696824"
      ]
     },
     "execution_count": 70,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.sqrt(mean_squared_error(y_test, y_pred_cell_line))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.1977430642893962, 1.7112782039203935e-28)"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scipy.stats.pearsonr(y_test, y_pred_cell_line)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.017526777020447904"
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f1_score(y_test, y_pred_cell_line)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.colorbar.Colorbar at 0x7f2d9b2d9400>"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f2d9b90d1d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist2d(y_test, y_pred_cell_line, (50, 50), cmap=plt.cm.jet, range=[[-10,10],[-10,10]] )\n",
    "plt.plot([-10, 10], [-10, 10], ls=\"--\", c=\"1\")\n",
    "plt.colorbar()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Based on the drug_id average"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_drug_id_mean = train_df.groupby(['drug_id'])['IC50'].mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_drug_id = np.array([train_drug_id_mean[drug_id] for drug_id in test_df[\"drug_id\"]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.823602003332211"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.sqrt(mean_squared_error(y_test, y_pred_drug_id))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8249253406877242, 0.0)"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scipy.stats.pearsonr(y_test, y_pred_drug_id)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6100478468899522"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f1_score(y_test, y_pred_drug_id)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.colorbar.Colorbar at 0x7f2d9b32a978>"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW8AAAD8CAYAAAC4uSVNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzt3Xl8VNX5+PHPk7CDbLIIogQsxQUrKiIq1ihIgVp3FNoqKBaxpZVfa79q/Vap1f5qa2utWv2hIriwuQEiFQHFtYKAioBYFkdFEETWyGbC8/vj3siYOTe5k8yduUme9+s1L2bOnHvuk0l4cnPuWURVMcYYU73k5ToAY4wx6bPkbYwx1ZAlb2OMqYYseRtjTDVkydsYY6ohS97GGFMNZSR5i8g4EdkkIsuSylqKyBwRWeX/2yLg2KF+nVUiMjQT8RhjTE2XqSvv8UD/MmU3APNUtQswz3/9LSLSErgFOBnoCdwSlOSNMcYckJHkraqvAlvKFJ8HTPCfTwDOdxz6A2COqm5R1a3AHFJ/CRhjjCmjToRtt1XVDQCqukFE2jjqHAp8mvR6nV+WQkRGACO8V3VPhFYZDdYYU3PUqZNH584t+Prr/Xz00YrNqtq6sm19R0R3hay7AWaralYuQKNM3mGIo8w5X19VxwJjAUTa6zd53BhjknTs2Jz584dxyCFNuOqqGXz00UUfV6W9XcDVIeuOyeJVZZSjTTaKSDsA/99NjjrrgMOSXncA1kcYkzGmhvvssx28+ean9O49jieeWFrl9gTvKjfMI5uiTN4zgNLRI0OB6Y46s4F+ItLCv1HZzy8zxpjQRITf/vY0WrVqRHHxfn7yk6dZvDgz14F5QMOQjxBxNhCRhSLynogsF5E/+OXjReQjEXnXf3SvqK2M/LIQkUlAIdBKRNbhjSD5MzBVRIYDnwCD/Lo9gJGqepWqbhGRPwJv+03dqqplb3waY0ygpk3r8/jjF/KjH3Vl374S7r77rYy2L0DdzDW3FzhLVYtEpC7wuoj823/vt6r6VNiGMpK8VXVIwFt9HHUXAVclvR4HjMtEHMaY2qVLl4OZPn0w3/lOS37+8+e5//63Kz4oTaXdJpmg3hrcRf7Luv6jUuty2wxLY0y1dMoph7Fw4c9o1aoRZ5/9WCSJGw5ceYd54PU+LEp6pIysEJF8EXkX7z7gHFVd4L91u4gsFZG7RKR+RXHlerSJMcZUyqpVX/LKKwmuvfYFPv54W2TnSfPKe7Oq9iivgqqWAN1FpDnwrIh0A24EPgfq4Y2qux64tbx27MrbGFNtNGhQhxtvPJ06dfLYvHkX558/OdLEDWlfeYemqtuA+UB/Vd2gnr3AI3gzzstlydsYUy0cdlgzXn/9Sm677SzOOqtT1s6b4dEmrf0rbkSkIdAXWJk0rFrwZqMvC27FY90mxpjYO+20w3n66Uto2LAu5547iRdfXJO1c2d4tEk7YIKI5OP9XpiqqjNF5CURae2f7l1gZEUNWfI2xsTa4MHdmDDhAhKJbRQWjmflys1ZjyGDo02WAsc7ys9Kty1L3saYWFu2bBMzZnzIVVfNYPv2PVk/f4avvDPG+ryNMbHTunVjfvnLkwEveQ8aNDUniRviOz3erryNMbHSvfshTJ8+hNatG/H88/9l7dqtOY2n9IZl3NiVtzEmNi69tBtvvDEcEejde1zOEzdEN1Swqix5G2Ni4fe/P4PJky9m8eL19OgxliVLNuQ6JMC6TYwxseLaLnaCowygIGRZx4Djwy2n/f77rbj//s+49toEX399UgXtus4P8Iqj7A+hzh8krjcsLXkbY3Kma9dGnHBCEyZN2sS0aZuZNi37wwArksmFqTIpjjEZY2qBgQNbMnHi0RQVlTBt2mZ2796f65Cc4nrlbX3expisu/76w3nuuWNZs2Y3p5yyJLaJG7zknanp8ZlkV97GmKx64omj+PGP2zJp0kaGD/8w1okb/CvvsJmyOMpIvs2StzG1UtDNSZdEyLJwVqzYz/XXL+Uvf3mj0m1kkwjUseRtjKmNTj/dGzHy2msfc/vtr+Y4mvSIQN38XEeRKtI+bxHpmrSh5rsiskNERpepUygi25Pq3BxlTMaY7Bo5sgfz5l3O7benvfZSLJReeYd5ZFOkp1PVD4Hu4G39A3wGPOuo+pqqnhNlLMaY7KpbN59//nMAI0f2YObM//KTnzyd65AqRQTqVrgpWfZl83dFH2CNqoYbsW+MqbYOOqg+zz//Y04/vSN/+tNr/P73L7F/f6X22c29mA70zmZIg4FJAe+dIiLvAeuB61R1efbCMsZkWlHRPj76aBv33ruQqVOr+X/n2py8RaQecC7eJptlLQE6qmqRiAwEpgFdHG2MAPydmJtFFqsxpvIuvbQbCxasI5HYxtChrh7SaiqGyTtbk3QGAEtUdWPZN1R1h6oW+c9nAXVFpJWj3lhV7eHtzNwo+oiNMaHl5Ql33HE2kydfzPXX9851OJklQH7IRxZl6/fJEAK6TETkEGCjqqqI9MT7hfJlluIyxlRR8+YNmDjxIgYM6MJ99y1k9OgXch1SZtXWbhMRaQScDVydVDYSQFUfAC4GrhGRYmA3MFhVq+mdDVO7FQSUJ7IYg0tByHqulQYDNBGv5Y4wexoUHA4/ewQeWtQTftrz23XHf+RoIGjcwvzwMWSLALVxtImq7gIOLlP2QNLze4F7o47DGJN5m76AtR/BFdfAmx1yHU1EYnrlbQtTGWPSdvWV0Lgx7NoFAy6EN9/KdUQRyuBuDCLSQEQWish7IrJcRP7gl3cSkQUiskpEpviDPMplydsYE1qjRjBlCjxwN1zx01xHk0WZu2G5FzhLVY/Dm8DYX0R6AXcAd6lqF2ArMLyihmL4x4AxpmIFadQNu2tO0I4z3vEdO9Zh+vQ2dOtWj+uuU+79W5lq41272ACcEfL84P66qnp8FWWw28S/n1fkvyzd+lKBs4Af++UTgDHA/eW1ZcnbGFOhXr3q89xzbalTB374w43Mnt021yFlT4b7vP2lQhYD3wHuA9YA21S1dE3CdcChFbVjydsYU6H160tYvnwfP/vZZlatyuK6p3GQ3miTViKyKOn1WFUdm1xBVUuA7iLSHG+tp6Mc7VQ44s6StzHGqV69fIYPP4EHHoBPPimmsPDzXIeUG+ldeW/2JhJWTFW3ich8oBfQXETq+FffHfCWCimXJW9jKqXAUZZI4/jCNOq6+nwd/ctHBrTpGMI3fE5RStn7nPnN85bU5Q6O4Xs04+p//ZxtLP5W3WNZmnL8ANyTcxZKwlFa4I61yhs/pFM3pAx2m4hIa+BrP3E3BPri3ax8GW/Oy2S8mwzTK2rLkrcx5luO4iD+wjEcRB1uZDn9yyTuWqd0enxmtAMm+P3eecBUVZ0pIiuAySJyG/AO8HBFDVnyNsZ8ow+tuYUj+ZK9/Ix3WMVX9M91ULmW2dEmS4HjHeVrgZ6pRwSz5G2M+caX7OMdtnEzH7A9mxsyxlltnR5vjIm3xtTnWDryPvAu27mW93MdUrzEdHp8DEMyNUtBQHkiizFkS2FAecgbjkGfSV9Hu3MddXsFnP6pgHKgPS0ZxUBa0ISrlnaGfZ+m1JEvHaPW7gxuM0UTR1lROrvX51hMk7dNjzemlupOJ37HxdSnLncyzZm4jS9Da5tkOiRjTC0zkBO5kFP4iI38i1ls5atchxRfmR1tkjGWvI2phYrYw5us5FFeppiSXIcTbzHtNolhSCZ3ChxliSq2WdXjo1IYsp6rv5pvNiMI5eKQ5y8IOD7hKFs2LKWo5zHuhaEue+QxAJrSiua05ROWc/l93jDiU5PqzR/lPr0OS/1apYOjH3yd+3i3WwLKgxbHyiEbbWKMyZUOHEl/RvI1+3iM3+U6nOrFrryNMbnwPfrQm0vYyuc8z73st/Hb6bHkbYzJpjzgJrryfc5gDUuYy8N8zd5ch1X91NbkLSIJYCdQAhSXXXFLRAS4GxgI7AKGqeqSqOMyLolcBxCBgoDyRLjDgxZ7Wjk+teyFYe66IcdEz3/kZGd54ewFKWX3HHNVSllf5qaUtWUMPJbgiCdu5Yjkfb3bh4sJQFyrk7rGjheND99odftZq8WjTc5U1c0B7w0AuviPk/F2j3D/FBtjKtSA41H2spcVbGQMLR4P2qndhBLTK+84TNI5D3hUPW/hrWvbLtdBGVMdNWMQhzOZNvxvrkOpOUpHm4R5ZFE2krcCL4rIYhEZ4Xj/UCB5apdzCyARGSEii7xdKnZFFKox1ZOQRxtuph1/ZTdv8xk/z3VINUcGd4/PpGyc7jRVXS8ibYA5IrJSVV9Net81YDalk83fSmgsgEj7CrcIMqa2qEdDBvBzWnIUW3iYTfwJbOJN5sS02yTykFR1vf/vJhF5Fm/N2uTkvQ44LOl1qC2ATLYUOMoSWY7BpSBk2fw0jnfssr7yI/fhRw4LaNfhH6lFw4+513F292JNw3+QWnfU+uS1+utAy4Hwxh20XDmBlmUrO/43bZndIKXs8Uf+6jx/g+1bUsr2rEs5C7w1zHk8Ra7PsMBd12l+GnUjENPp8ZF2m4hIYxE5qPQ50A9YVqbaDOBy8fQCtqvqhijjMqZGqP8jyGsNFMOWgbCyGq3UV53U0m6TtsCz3mhA6gATVfUFERkJoKoPALPwhgmuxuvMviLimIyp1gTgoDFw0C1Q9E/YcW2OI6rhBEj9QyXnIk3e/tY+xznKH0h6rsAvoozDmJqiPnW5irPhoM6w6xHY8T+5Dqnmi2m3SQy74U28JLJ4rsI0zh+wYFQKRz82wLBOqWWuhZU6OOoBDf6R2g/8vWbuHWiWbj82pWwjbVPKrguYzTPKX0SKgzrCD2dCi8P59+hfseCee75V7xZHNzRAIjVUOt23J6Vs7i/6Oo/f8w9X/7ajYtEYdwBOBQHliTTayJLaesPSGJMh+3ZC8W547gcsuOelXEdTu8QwU8YwJGPMt3QZAmuegr1b4Km0Nhg3mRDTbpM4zLA0xjjkU5e+XAn9JqY3NNFkVgZHm4jIYSLysoh8ICLLReRav3yMiHwmIu/6j4EVtWVX3rFXkEbdRJbOH8V5IL3xvGE39Q3o8x7vGHu8zNG/7erbBa5rlto/3Zd5zrpnvLIwpezJc89JKRs0Y+aBFw3aw0nPQoueMOH3MOmhb01dK9EbU46/2R0q33PsBn87P0gp+/iKI90NuD4r1z2D8elssJAIqOtSUMXjqyizmzEUA79R1SX+MOrFIjLHf+8uVQ29tbMlb2PipnlP6DkN6jSBhefBxBm5jqh2y+ANS38Oywb/+U4R+QDHciBhWLeJMXFT8hXs/Rxe6wWfW+LOufS6TVqVrsHkP1zrOXnNihQAxwOla/6OEpGlIjJORFpUFJYlb2NiQMiH9pd6L3Yuh1dOgJ0rchuU8aSXvDerao+kx1hnkyJNgKeB0aq6A28p7COA7nhX5n+rKCxL3sbkWD2a8n1uhR6TocUpuQ7HuOSHfIQgInXxEvcTqvoMgKpuVNUSVd0PPIi3BlS5rM879hLV6PwFaRxfWMVzOaS1WJTjhptrd5gO7sPv3H5dSlnfZu4bliefOz+lbOHZ3g3X73WGaWOg6cFw2fPw+NI3U+qeM/vJlLK52/uklLluogLcSWqsrolDwbu/F6QWjXct7Bm087vj+LS+1+nUjUAG+7z9ncMeBj5Q1b8nlbdLWtPpAlLXgEphyduYHLmgNzz2P7C1CE7/P7CoTa4jMk6ZHW1yGnAZ8L6IvOuX/Q4YIiLd8cYUJYCrK2rIkrcxOVJSAu+sgYtvhY1bAUve8ZTZ0Sav497DYFa6bVnyNiaLGpNPd5qxEJjxH3juLVDbWiTebG0TY5IlQpYFTLJxcfRPN3jKsSoTsKeXY7GlXqlFHW9Z6TzeNaHlzkdS+5bhQP/yEXl5jG90EAV5eUxcCxs3ldkyarDzcGaePSi10DF56Laifs7jezp+OTg3UwiUcJSls6mx6/hqxJK3MbVX3zp1ebRRY0oUzv9qJxs3Nc11SCYNGsO1TSx5GxOxa+s14PYGDVm+v4RBXxXxie7PdUgmDZoH+2rbZgzGGGggMO3rfYzY/RW7ch2MSZsKFOeHnRKTvV/MlrxNkoKQZfOr2GaAJmNSy4qqdjevbbNNzvLrPkjdgWa2Y7GmoA0SzvhF6mJTv+fAYlFNaUYzmvIpn3JMg0fZDzxQ70DdjTtTN2PYGDDc5LbX/5Ra6PpclhU6j18YdmPBgEW43P3bHVOLjhzjPnzleEdhIuBchY6y+QF1s0NFKKkTNlXuizSWZJHNsAxa+rBMnUIR2Z60DGLQwmjGVBuHczjX8Asu5lLybBJzjVCSnx/qkU1RXnk7lz5U1bILNrymqqnrYxpTDZ1ID87hXLaxjYk8xv4s/hltoqEIJTHcjSGy5F3O0oe22o6pgfI5h3M5mV6s4r9MZTJ7SN0n0lQ/ilBcm5J3MsfSh8lOEZH3gPXAdaq6PKCNEYC/vGKzKMI0pgpKaEQjXuNV5jAbxWbe1BSKsC+D8+MzRTTi6V3+0oevALeXrqCV9F5TYL+qFvnb/tytql0qbrO9fpPHjcmh4447hJ0797J27Vby8oT9+y1px88fFqtqj8oefWyPejpjUatQdTvLhiqdKx2R3k1xLX2YTFV3qGqR/3wWUFdEwn1KxuTYJZccw5tvDufee73tBi1x11wl5Id6ZFOUo02cSx+WqXOIXw8R6enH82VUMRmTCXl5wu2392HKlEEsWbKBYcOm5TokE6HSPu8wj2yKss87aOnDwwFU9QHgYuAaESkGdgODNep+HGOq4KCD6jNx4kWcc853GTt2MaNGzeLrr0tyHZaJkDfaJH5TYqIcbRK09GFynXuBe6OKwQQpCChPVLHdwtSiJo4ycE8yaeL4cXFtkBDENcnk9YC6AYtApZj87ZfF9aH1EXDNH+CBySfCESceeNOxsBUA4+enlg0rDB+Ta5OEyY6yoM0UXOVF4x2FZwQ0MCGgvHbwbljWq7hilsXv14kxMXRmd3j7QyjaDaeOhv2f5Doiky0KsRwqaNO/jKnA/1wKc++Am3/qvd5v825qGa/bJMwjm+zK25gADfPh4d/BkDNh8stwy6O5jsjkQq2bYWlMddahMUzvC91bwg0PwR1Tch2RySVL3qYSChxliSq2mc7xVTx/4KqAjp3GiwpTy/oHNeyIockwR5vj3Ye/5aibdMNRWkOT3vCju2HWUlJ36XHdHA28uZpILXLdxBzvWKkP8Oa4VYXj/M7va1XPUzPZlbcx1cDAnvDvt+HTL+Do4VDSPtcRmVxThL0xnB5vNyyNAerWhfuvhedvh5/28cpK7Mak4cCVdyZmWAYtlS0iLUVkjois8v9tUVFblrxNrdemNcybCSPPgf87CZ54KdcRmTjJZPLmwFLZR+F11P1CRI4GbgDm+Ws7zfNfl8u6TWIvUY3OP99RFtCP69p1xbXjyrBh7uNdE096O8rGBEw8udj75/iOMO2X0KoJDBmyicmTvypTscB9/JGOCUXOHWNw98Vf7Kg3PuD4SH4GXG0WRHCemiFT47zLWSr7PA7McpuA95/p+vLasuRtarWGdWFvMfT+v/BOSuI2Ju3p8a1EZFHS67GqOtZVscxS2W39xI6qbhAR9554SSx5m1onLw/6doMXl8Gbq+Hom6DYlicxAdIcbbI5zJKw/lLZTwOjVXWHvz5fWqzP29QqzZrlMXNmW2b/Bk7we3QscZvyeKNN6oV6hBGwVPZGEWnnv98OcO+cncSuvE2SAkdZImS9oLoBixqtLKwwGqCcfmCHua7+7QNjl7t2bcaMGWfRqVMDrr76OZaMXVymbkG5x3/LyoSjsNBd1zXWfPxQd10nV7uu8wctLOX6GlzHu8pMJlcVLGep7BnAUODP/r/TK2rLkrepFQYO7MDEid9n794SzjrrBV5/vWziNiZYBifpBC2V/WdgqogMBz4BBlXUkCVvUyscckhDVq/ewQUXvMynn9qNSRNeJmdYVrBUdp902rI+b1NjNWoknHaad9N+3LhV9Or1vCVuk7YMj/POGLvyNjVSx451mDatDUcccSgFBU+xZcteiottkyaTvrhOj7fkbZIkqnh8gaMs6Caai2tCz8cBdV3teotdff/7HXnqqUuoW1e55JIpbNmy2lHXdcMw7I09SG/BLldd143coONd5WHbNFUV14WpIu82EZH+IvKhiKwWkZQpnyJSX0Sm+O8v8AeuG1Mp11xzEnPnXs7mzbvo2fNBXnjBlbiNSU8cu00iTd4ikg/cBwwAjgaG+PP4kw0Htqrqd4C7gDuijMnUbF27Hszs2Wvo1eshVq36MtfhmBqgNu4eD9ATWK2qawFEZDLeHP4VSXXOA8b4z58C7hURsV3kTVht2kDr1rB8OfzmNy+iquzfbz8+JjNq3e7xvkOBT5NerwNODqqjqsUish04GNicXElERgAjvFfNook2awoCyhNVPL78SSoVH+/oc27SKbUsaIMF1+7vQa5LLWowektK2Z6LC93H+5shnHg8PDsRir6CbneMoaRsaEEbJLgWhnrK8bWW3YShlGthLNcGDeDeVd75GTo2qAiUcJQVpHF82DajUhCDGMKrjX3erv/NZX9qw9RBVceqag9v3YBGGQnOVG8/uRRemw0lJTDkCrCLbRMFRdhHvVCPbIo6ea8DDkt63QFYH1RHROrgXVanXoIZ48vLg7/cBo8/BAsWwUmF8N77uY7K1FRx7fOOOnm/DXQRkU4iUg8YjDeHP1npnH7w/ph9yfq7TXkEOP57cM8DcPa5sHlzhYcYU2mlfd5hHtkU6dn8PuxRwGwgHxinqstF5FZgkarOwFuk5TERWY13xT04ypjiIVHF4wsCyl39267xzEHnd4ypLgo6l4Orf9jVNwwwJvX3855Ey3KPP+oI+HIbbPoSfvhX2Pc1kLz4pmszhqDzj3ddHzjGSa9MYwGpbomAN9IZP+5S6Cibn8bx6ZwrWxK5DiAtcezzjvxXharOAmaVKbs56fkeQizCYmq3H50JT/wVXnwDLr7WT9zGZEGtnaRjTFXdNBJm/AtWroVr/5TraExtE9c+7/gNXjTG17g+PHIXDOoPj82AETfDnr25jsrUNt5oE1vbxJjQ6tWB73WF39wBfx+f62hMbRXXbhNL3jlRkEbdhKNsfkDdwjTjKMsxyWeYYxh+0GQU1w3LoEkugx3t+jcXTz4U3vkctpbAcX/yNghOmejimnzjumE5N2jgUiKgvKygxZ7CHg/u73dhBG2aqMQxeVuft4mVUT3h9Svhd6d7r/cW5zYeY6zP25hy1MuDf50Lw4+HaSvhzjdzHZExntq6tokxFTqkITzTF05pA7e+AmPmO9ZHMCZHSqfHx40l75xIRNTu/JBlaRifRt2VadSde+Bpy6NbU3Dm5Vx00SyeeeaDNBopY0zlD41WImRZOsebbCntNokbS94mZ0499TDefPNTVqz4gs6d72bPHuvgNvEUx24Tu2Fpsi4/P4+///0HvPHGcC66yNubwxK3iatMbkAsIuNEZJOILEsqGyMin4nIu/5jYJi44vfrxNRoLVs2ZMqUQfTt25l//OMtpk1Lp6/FmOzL8Djv8cC9wKNlyu9S1TvTaciSd+wVOsoSWY6hrKBNhV2bBSe+eXbMMc2ZPv0sOnRozBVXrGT8+D3A95PqznccXxBwrkRAedjjqyrs+YMURNCmiUqm+rxV9dVM7dNrydtkTadOTWjQIJ8zzniBBQsa5zocY0LZT1460+NbiciipNdjVXVsiONGicjlwCLgN6q6taIDLHmbSInASSe1YuHCzcycuY4uXZ5h9+4SwJK3qT7S6DbZ7O32lZb7gT/ijZD9I/A34MqKDrIbliYyTZrk89RTZ/LGGwM5+ujmAH7iNqb6yOQNS2f7qhtVtURV9wMP4m3cXiG78jaR6Ny5AdOnH8uRRzbkuusWsWLFtlyHZEylKJnr83YRkXaqusF/eQGwrLz6pSx5x978CNosqGJd184wB+r26dOCqVO9IYD9+29k3rzWpNzkdO5Kn3C06SoLiiud47OpINcBmCrJ3PR4EZmENwqhlYisA24BCkWkO97viQRwdZi2LHmbjDvppIP47LO9nH/+MtaubZvrcIypkkwOFVTVIY7ihyvTViTJW0T+CvwI2AesAa5Q1ZS/m0UkAewESoDiSnT0m5ioXz+frl0bs3TpV/z5z59w993r2L17f67DMqbKFGFvDNc2ieqG5Rygm6p+D/gvcGM5dc9U1e6WuKuv9u0b8cor/Zk3rztNm3pXKJa4TU1Rq3aPV9UXk16+BVwcxXlypyCgPBHD87vqBk2ycfRlH1mYWrbyo2+e9upVn2eeaUOTJnlcdtlUduwoO2PSdX6gKKjfPKxEyHoB5w/9WYU9T3ky0YbJpdq6GcOVwL8D3lPgRRFZLCIjymtEREaIyCJvAPyujAdp0jdsWBPmz2/Hrl3KKaesZ/p0m+puap6ohwpWVqWvvEVkLnCI462bVHW6X+cmoBh4IqCZ01R1vYi0AeaIyEpVfdVV0Z+lNNZrt70t9xwD/fo15NVX93DppZvYutW6SUzNpAgl++N35V3p5K2qfct7X0SGAucAfVTVmWxVdb3/7yYReRZvcLozeZt4OLg5NOlYh48/LubKKzfz9ddKic27MTWY7hf27onf7vGRdJuISH/geuBcVXX2cYhIYxE5qPQ50I+Qg9NNbhz7XXj7SXj66TYA7NljidvUfKpCSXF+qEc2RXV79F6gPl5XCMBbqjpSRNoDD6nqQKAt8Kz/fh1goqq+EFE8GZYIKC8IWddVL4jj5uKRjgku8M3u65XWwdGuf3Pywgsb8eijrdm2bT/XXFOP8DuiB90c/UPI44emcfwtjrKgG6OJkGUFaRyfTt0o5Pr8NZSS9cQcRlSjTb4TUL4eGOg/XwscF8X5TeaIwJgxzbn55hb85z97uPDCTXz++WG5DsuYrFEVir+uJcnb1Bz16ws//GEjxo3byTXXbGbfvlxHZEy2CftL4pcq4xeRiYXOh8EXW2DnHqWwcANFRTbAx9RSCtSWbpOaryCgPKh/N+TxTQpTy4rmp5YFDad2LvaURtL12+3XDyZPhhkOnzOWAAASs0lEQVQzYNiwAoqKylacENCAq386nck4HR1lQedycfWDByl0lM13lCXSaDOdulHI9flrqP0Ce+KXKm09b/Mtv/41zJoFn3wCY8bkOhpjYqI45COL4vfrxOREgwYwdixcdpnw1FPKsGHw1Ve5jsqYGPAW9I4du/I2ALRsCX36wP/+rzJokCVuY75RmrztyrsmSKRZHrKeq387HSl90xU75pg2rFjxBevXK0ceWZ+dO/dWIYB0+pxd0unfrqr5WTyXqdYU+DrXQaSyK+9abPjwE1iy5Gp+/etTAKqYuI2poRTYG/KRRXblXQvVqZPHXXf1Z9SonsyevZqHH16S65CMia+Y9nlb8q5lWrVqxNSpgzjzzE789a9vcOON8ygpsRUBjQlkydvEQefOLTj++Hb89KfP8MQTS3MdjjHxZ8m7NigIWS9gsaUjJbUsadeaA/UCFqZy1fVj6tYNli2DhQuhoAC2b78AuCBErAlH2cdp1E1j1x6ngvDnajIstSxwklLYm6tpnN/UTDFN3nbDsoYTgdtug/ffF/r398q2b89tTMZUOxkaKigi40Rkk4gsSyprKSJzRGSV/2+LMCFZ8q7BmjYVpk+Hm24SHnxQeemlXEdkTDW0H9gT8lGx8UD/MmU3APNUtQswz39dIUveNVSXLnV466329O8PP/+5MmIEtiKgMZWRwUk6/jaPW8oUn8eBSQ4TgPPDhGV93hmVCFkvoL817P69Ier16nUcrVr14+yzJ/LKK2HjikJVz53G8UVjqniuKp7f1Ezp9Xm38jZJ/8ZYf//d8rRV1Q0AqrrB39O3Qpa8a5ijjmrNBx98wWOPvcdzz33Itm3h/pYzxgRIL3lvVtUe0QVzgHWb1BANGtTh8ccvZPHiERxxREsAS9zGZEq0a5tsFJF2AP6/m8IcFFnyFpExIvKZiLzrPwYG1OsvIh+KyGoRCdVRb76tQ4emvP76lQwZciy33voKa9aU7VIzxlRa9AtTzeDA+OGhwPQwB0XdbXKXqt4Z9KaI5AP3AWfjbZ/7tojMUNUVEcdVRUGb4oZdWKkwQ3HAaac14+mnj6FhwzzOPXcezz+/jm+PTU44jipwlKVT11UvE7J5LmNC2g/szkxTIjIJLwG0EpF1eLtm/xmYKiLDgU+AQWHaynWfd09gtb8ZMSIyGe/Oa8yTd3ycf34rtm8vprBwGStXVnX7eGNMCgVKMtSU6pCAt/qk21bUfd6jRGSpPzDdNfD8UODTpNfr/LIUIjJCRBZ5d3J3RRFrtVG3rtC5cwMAbrhhLSedtJiVK2v3Z2JMpGK4nneVkreIzBWRZY7HecD9wBFAd2AD8DdXE44y53xmVR2rqj28O7mNqhJ2tda6dV3mzj2O+fO706hRHiUlyo4dGbosMMakqombMahq3zD1RORBYKbjrXXAYUmvOwDrqxJTTda9exOmT+9G69Z1ufLKD9m1y1YDNCZyMV3bJLI+bxFpVzrwHG8FpGWOam8DXUSkE/AZMBj4cVQxZU5Vd3yZn/YRl17ajXHjTmPz5l307j2BJUs2VHxQoEREdasqm+cyJqTS6fExE+UNy7+ISHe831sJ4GoAEWkPPKSqA1W1WERGAbOBfGCcqi6PMKZq64orurN48XouumgqX3xhG0wak1W16cpbVS8LKF8PDEx6PQuYFVUc1VnTpvWpX78OX3zxFZdc8iS7dxfz9dfWv21MVsW028RmWMbUd797MAsW/IypU70hnzt27LXEbUwulG5AHOaRRbke520cBgzowqRJF7F3bwkjRjyX63CMqd0yOM47k+zKO2auv743M2f+mDVrttKjx1heey1o1xpjTFbUxKGCJrOaNKnH8OHHM2XKMoYPn8Hu3Vn+O8wYk0rJ2PT4TLLkHQMdOjTliy92UVS0j1NPfZjNm222pDGxYd0mxuX00zuyZMnV3HlnPwBL3MbETUy7TSx559DIkT2YN+9yvvxyN//854Jch2OMcYlp8rZukxyoWzefe+4ZwNVX92DmzP/yk588zY4de3MdljHGpXSoYMzYlXcOHH54MwYP7saf/vQa5503yRK3MXFXEvKRRXblnUUFBc1JJLaxZs0Wuna9l40bi3IdkjGmIjFd28SuvLPkxz8+lhUrfsHPfnYigCVuY6qLmM6wtOQdsbw84Y47zuaJJy5i4cLPePbZD3IdkjEmHaVDBa3bpPZo3rwBEydexIABXbjvvoWMHv0CxcW2Brcx1U4MF6ay5B2hHj3ac+aZnRgx4jkefHBxrsMxxlRGTFcVtOQdgcMPb8Ynn2xn7ty1dOr0Dz7/3Pq3jam27IZl7XDjjaezevWv6N37cABL3MZUdxmepCMiCRF5X0Te9TZUrxy78s6QRo3q8sgj53PJJcfw+ONLWbTItuI0psbIfLfJmaq6uSoNWPLOgI4dmzN9+mC6dWvDdde9yN/+9mauQzLGZEpMZ1hGkrxFZArQ1X/ZHNimqt0d9RLATrxBNsWq2iOKeKJ23nldOfzwZgwc+AQvvrgm1+EYYzIp86sKKvCiiCjw/1R1bGUaiSR5q+qlpc9F5G/A9nKqV/nPh1zp0KEp69bt4J//XMCTT65gw4aduQ7JGJNp6Y02aVWmH3usIzmfpqrrRaQNMEdEVqrqq+mGFWm3iYgIcAlwVpTnybZ69fK5774fctFFR3HccQ/w6afbLXEbU1PtJ53NGDZX1IPgb8KOqm4SkWeBnkDayTvq0SanAxtVdVXA+6V/PiwWkRERx5IRbds24eWXh3HVVSdw331vs27djlyHZIyJWoZmWIpIYxE5qPQ50A9YVpmQKn3lLSJzgUMcb92kqtP950OASeU0E/rPBz+5+wm+WWXDrpIePdrz7LODadGiAYMGTeWpp1bkJA5jTJZpxlpqCzzrdUpQB5ioqi9UpqFKJ29V7Vve+yJSB7gQOLGcNkL/+eD3G4312m6fuY8yDSNH9qC4eD+nnvowS5duzEUIxphqTFXXAsdloq0o+7z7AitVdZ3rTf9PhjxV3Zn058OtEcZTKfn5ebRu3YjPPy9i1KhZNG5cjy+/tK3KjDG5FWWf92DKdJmISHsRmeW/bAu8LiLvAQuB5yv750NUWrRoyKxZP+Gll4ZSv34d9uwptsRtjImFyK68VXWYo2w9MNB/nrE/H6Jw9NGtmT59CIcf3oxrrpnJ3r0xXJnGGJMF6Q03yRabYelw3nlH8thjF1BUtI8zzniEt95y9vwYY2qFeE6xtORdhohw4429+eCDzVxwwWTWr7fx28bUbvFcE9aSt69x43rk5Qk7d+7l3HMnsX37XusqMcYQ1ytvWxIW6NSpBf/5z3AeffQCADZt+soStzHGF89NLGv9lfdZZ3Vi6tRBiAi//vXsXIdjjIkdJY43LGv1lfcvf3kys2dfxoYNRZx00ljmzl2b65CMMbGT4d0YMqTWXnm3aNGQG2/szXPPfcjllz9LUdG+XIdkjImlePZ517rk3bp1Y778chdbt+6mV6+H+PTTHajmZLa9MaZaiOdok1rVbdKz56G8++5IxowpBOCTT7Zb4jbGVCCeNyxrTfIeOrQ7r756BXv2FDN16vJch2OMqTaszzsn8vPzuPPOfowe3Yu5c9dy6aVPsmVL/O4cG2PiyqbH58RRR7Vi5Mge3HXXf/jtb+dQUrI/1yEZY6oVu2GZVW3aNGbTpq9YtmwTRx99Hx99tDXXIRljqi27YZkVF154FGvWXMugQccAWOI2xlSB3bCMnIgwZkwhTz99KcuWbeL11z/JdUjGmGovnsm7xnSbNGlSj8ceu5Dzzz+SRx55h2uued7WJzHGZEA8x3nXmOTdr98RnHPOd/nVr/7NPfcsyHU4xpgaw0abRKJVq0Zs3ryLZ575gKOOupfVq7fkOiRjTI0Sz9Em1brPe/ToXnz00Wi6dz8EwBK3MSYC8ZykU6XkLSKDRGS5iOwXkR5l3rtRRFaLyIci8oOA4zuJyAIRWSUiU0SkXsjzMmHCBdx1V39mz17NqlWWtI0xUcnsDUsR6e/nxdUickNlo6rqlfcy4ELg1TLBHY23e/wxQH/gXyKS7zj+DuAuVe0CbAWGhzlp166tuPzy4/j9719i0KAn+eorWxHQGBOVzF15+3nwPmAAcDQwxM+XaatS8lbVD1T1Q8db5wGTVXWvqn4ErAZ6JlcQEQHOAp7yiyYA54c5b8OGdTj//MncdturtrCUMSZipTcswzwq1BNYraprVXUfMBkvX6YtqhuWhwJvJb1e55clOxjYpqrF5dT5hoiMAEb4L/e+886QZRmKtbJaAZtzHAPEI444xADxiCMOMUA84ohDDABdq3b4htkwplXIyg1EZFHS67GqOjbp9aHAp0mv1wEnVyaqCpO3iMwFDnG8dZOqTg86zFFW9hI5TJ0Db3gfwFg/pkWq2iOobjbEIYa4xBGHGOISRxxiiEsccYihNI6qHK+q/TMVC2nmvfJUmLxVtW8l2l0HHJb0ugOwvkydzUBzEanjX3276hhjTE0SJjeGEtVQwRnAYBGpLyKdgC7AwuQK6nVWvwxc7BcNBYKu5I0xpiZ4G+jij7SrhzewY0ZlGqrqUMELRGQdcArwvIjMBlDV5cBUYAXwAvALVS3xj5klIu39Jq4Hfi0iq/H6wB8OeeqxFVeJXBxigHjEEYcYIB5xxCEGiEcccYgB4hMHfi/DKGA28AEw1c+XaRMbrWGMMdVPtZ5haYwxtZUlb2OMqYZim7xzNfW+nHimiMi7/iMhIu8G1EuIyPt+vSoNUQpof4yIfJYUy8CAehmZghvQ9l9FZKWILBWRZ0WkeUC9jH8WFX1d/k3yKf77C0SkIBPnLXOOw0TkZRH5wP8ZvdZRp1BEtid9n26OII5yP1/x/NP/LJaKyAkRxNA16Wt8V0R2iMjoMnUi+SxEZJyIbBKRZUllLUVkjv//fo6ItAg4dqhfZ5WIDM1EPFmnqrF8AEfhDa6fD/RIKj8aeA+oD3QC1gD5juOnAoP95w8A12Qwtr8BNwe8lwBaRfi5jAGuq6BOvv+5dAbq+Z/X0RmMoR9Qx39+B3BHNj6LMF8X8HPgAf/5YGBKBN+DdsAJ/vODgP864igEZkb1cxDm8wUGAv/GG1vcC1gQcTz5wOdAx2x8FsD3gROAZUllfwFu8J/f4PrZBFoCa/1/W/jPW0T52UTxiO2Vt+Zo6n1F/LYvASZlor2IZGwKrouqvqgHZsa+hTdWNRvCfF3n4X2/wfv+9/G/ZxmjqhtUdYn/fCfeqIHA2cE5dB7wqHrewptX0S7C8/UB1qjqxxGe4xuq+ipQdlW65O9/0P/7HwBzVHWLqm4F5uCtwVStxDZ5l8M1vbRKU+/TdDqwUVVXBbyvwIsistif0h+FUf6fweMC/iwM8xllypV4V3cumf4swnxd39Txv//b8X4eIuF3yxwPuHYAOUVE3hORf4vIMRGcvqLPN5s/B+D9pRN0URP1Z1GqrapuAO+XLNDGUSfbn0skcroZg8Rk6n2a8Qyh/Kvu01R1vYi0AeaIyEr/CiG08uIA7gf+iPf1/BGvC+fKsk04jk1rTGiYz0JEbsJbSu2JgGaq/FmUDctRlpHvfWWISBPgaWC0qu4o8/YSvO6DIv++xDS8yWqZVNHnm83Poh5wLnCj4+1sfBbpyNrnEqWcJm+N2dT7iuIRkTp4S+CeWE4b6/1/N4nIs3h/6qeVsMJ+LiLyIDDT8VaVp+CG+CyGAucAfdTvSHS0UeXPoowwX1dpnXX+96sZqX9aV5mI1MVL3E+o6jNl309O5qo6S0T+JSKtVDVjCzWF+HwzNhU7hAHAElXd6Igz8s8iyUYRaaeqG/wuok2OOuvw+uFLdcC7t1atVMduk1xOve8LrFTVda43RaSxiBxU+hzvxl5GVz8s02d5QUD7GZuCGxBDf7zZseeq6q6AOlF8FmG+rhl432/wvv8vBf1yqSy/D/1h4ANV/XtAnUNK+9pFpCfe/7UvMxhDmM93BnC5P+qkF7C9tEshAoF/kUb9WZSR/P0P+n8/G+gnIi38bsd+fln1kus7pkEPvMS0DtgLbARmJ713E96ogw+BAUnls4D2/vPOeEl9NfAkUD8DMY0HRpYpaw/MSjrne/5jOV4XQ6Y/l8eA94GleD+o7crG4b8eiDcKYk2m4/A/00+Bd/3HA2VjiOqzcH1dwK14v0gAGvjf79X+979zBN+D3nh/Zi9N+gwGAiNLfz7wpkAv97/+t4BTMxyD8/MtE4PgLfy/xv+Z6ZHJGJJiaYSXjJsllUX+WeD9stiAt4XNOrzNXA4G5gGr/H9b+nV7AA8lHXul/zOyGrgiis8l6odNjzfGmGqoOnabGGNMrWfJ2xhjqiFL3sYYUw1Z8jbGmGrIkrcxxlRDlryNMaYasuRtjDHV0P8Hv2pmQT5KywkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f2d9b90e518>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist2d(y_test, y_pred_drug_id, (50, 50), cmap=plt.cm.jet, range=[[-10,10],[-10,10]] )\n",
    "plt.plot([-10, 10], [-10, 10], ls=\"--\", c=\"1\")\n",
    "plt.colorbar()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Based on the drug_id average and cell_line average"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/soeren/anaconda3/lib/python3.6/site-packages/pandas/core/indexing.py:357: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[key] = _infer_fill_value(value)\n",
      "/home/soeren/anaconda3/lib/python3.6/site-packages/pandas/core/indexing.py:537: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  self.obj[item] = s\n"
     ]
    }
   ],
   "source": [
    "train_df.loc[:,\"drug_id_average\"] = [train_drug_id_mean[drug_id] for drug_id in train_df[\"drug_id\"]]\n",
    "train_df.loc[:,\"cell_line_average\"] = [train_cell_line_mean[cell_line] for cell_line in train_df[\"cell_line\"]]\n",
    "\n",
    "test_df.loc[:,\"drug_id_average\"] = [train_drug_id_mean[drug_id] for drug_id in test_df[\"drug_id\"]]\n",
    "test_df.loc[:,\"cell_line_average\"] = [train_cell_line_mean[cell_line] for cell_line in test_df[\"cell_line\"]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = train_df[['drug_id_average', 'cell_line_average']]\n",
    "y_train = train_df['IC50']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "reg = LinearRegression().fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = np.array(reg.predict(test_df[['drug_id_average', 'cell_line_average']]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6690307328605201"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f1_score(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.730882767376305"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.sqrt(mean_squared_error(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8467616723142585, 0.0)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scipy.stats.pearsonr(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "|                                                | RMSE  | rPearson | F1 score |\n",
    "|------------------------------------------------|-------|----------|---|\n",
    "| pairwiseMKL                                    | 1.682 | 0.858    |   |\n",
    "| KronRLS-MKL                                    | 1.899 | 0.849    |   |\n",
    "| cell_line average                              | 3.201 | 0.196    |   |\n",
    "| drug_id average                                | 1.926 | 0.807    |   |\n",
    "| weighted drug_id average and cell_line average | 1.765 | 0.841    |   |\n",
    "|"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.colorbar.Colorbar at 0x7f2d9b4686d8>"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f2d9b3d85f8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist2d(y_test, y_pred, (50, 50), cmap=plt.cm.jet, range=[[-10,10],[-10,10]] )\n",
    "plt.plot([-10, 10], [-10, 10], ls=\"--\", c=\"1\")\n",
    "plt.colorbar()"
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}