IsolationForest.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-06T16:57:57.705221Z",
     "start_time": "2020-05-06T16:57:57.700078Z"
    }
   },
   "source": [
    "<center><h2>Isolation Forest</h2>From scratch</center>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "#### Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:28.727244Z",
     "start_time": "2020-06-14T17:32:28.714040Z"
    },
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "utils.load_extension('collapsible_headings/main')\n",
       "utils.load_extension('hide_input/main')\n",
       "utils.load_extension('autosavetime/main')\n",
       "utils.load_extension('execute_time/ExecuteTime')\n",
       "utils.load_extension('code_prettify/code_prettify')\n",
       "utils.load_extension('scroll_down/main')\n",
       "utils.load_extension('jupyter-js-widgets/extension')\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%javascript\n",
    "utils.load_extension('collapsible_headings/main')\n",
    "utils.load_extension('hide_input/main')\n",
    "utils.load_extension('autosavetime/main')\n",
    "utils.load_extension('execute_time/ExecuteTime')\n",
    "utils.load_extension('code_prettify/code_prettify')\n",
    "utils.load_extension('scroll_down/main')\n",
    "utils.load_extension('jupyter-js-widgets/extension')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:35:44.695378Z",
     "start_time": "2020-06-14T17:35:44.687956Z"
    },
    "hidden": true
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import random\n",
    "from pprint import pprint\n",
    "import seaborn as sns\n",
    "import pdb\n",
    "sns.set_style(style=\"whitegrid\")\n",
    "from matplotlib import rcParams\n",
    "\n",
    "%matplotlib inline\n",
    "plt.style.use('fivethirtyeight')\n",
    "rcParams['axes.labelsize'] = 14\n",
    "rcParams['xtick.labelsize'] = 12\n",
    "rcParams['ytick.labelsize'] = 12\n",
    "rcParams['text.color'] = 'K'\n",
    "rcParams['figure.figsize'] = 16,8"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sample data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:30.704915Z",
     "start_time": "2020-06-14T17:32:30.289608Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 504x504 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "mean = [0, 0]\n",
    "cov = [[1, 0], [0, 1]]  # diagonal covariance\n",
    "Nobjs = 2000\n",
    "x, y = np.random.multivariate_normal(mean, cov, Nobjs).T\n",
    "#Add manual outlier\n",
    "x[0]=3.3\n",
    "y[0]=3.3\n",
    "X=np.array([x,y]).T\n",
    "X = pd.DataFrame(X,columns=['feat1','feat2'])\n",
    "plt.figure(figsize=(7,7))\n",
    "plt.plot(x,y,'bo');\n",
    "plt.savefig('images/sample.png')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Isolation tree\n",
    " 1. Randomly select dimension\n",
    " 2. Randomly pick a value in the dimension\n",
    " 3. Split the data\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"images/tree_pseudo.png\" alt=\"Drawing\" style=\"width: 400px;\"/>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-06T16:05:41.616623Z",
     "start_time": "2020-05-06T16:05:41.612745Z"
    }
   },
   "source": [
    "### Select a dimension"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:30.734640Z",
     "start_time": "2020-06-14T17:32:30.710236Z"
    }
   },
   "outputs": [],
   "source": [
    "def select_feature(data): \n",
    "    '''\n",
    "    Randomly select a feature of a dataframe\n",
    "    '''\n",
    "    return random.choice(data.columns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Select a value in the dimension"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:30.751099Z",
     "start_time": "2020-06-14T17:32:30.745261Z"
    }
   },
   "outputs": [],
   "source": [
    "def select_value(data,feat):\n",
    "    '''\n",
    "    Select values of \n",
    "    '''\n",
    "    mini = data[feat].min()\n",
    "    maxi = data[feat].max()\n",
    "    return (maxi-mini)*np.random.random()+mini\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:47.593388Z",
     "start_time": "2020-06-14T17:32:47.586090Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "feat2\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "-2.427891971028951"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "var = select_feature(X) \n",
    "print(var)\n",
    "select_value(X,var)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Split Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:48.284258Z",
     "start_time": "2020-06-14T17:32:48.280353Z"
    }
   },
   "outputs": [],
   "source": [
    "def split_data(data, split_column, split_value):\n",
    "    '''\n",
    "    Split data based on the value of a column\n",
    "    '''\n",
    "    data_below = data[data[split_column] <= split_value]\n",
    "    data_above = data[data[split_column] >  split_value]\n",
    "    \n",
    "    return data_below, data_above"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:48.592747Z",
     "start_time": "2020-06-14T17:32:48.566769Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "feat2 -2.7308562661821796\n"
     ]
    }
   ],
   "source": [
    "var = select_feature(X) \n",
    "value = select_value(X,select_feature(X))\n",
    "print(var,value)\n",
    "a,b =split_data(X,\n",
    "           var,\n",
    "           value)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:48.853178Z",
     "start_time": "2020-06-14T17:32:48.846914Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(6, 2)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:49.185625Z",
     "start_time": "2020-06-14T17:32:49.179979Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1994, 2)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "b.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:49.583838Z",
     "start_time": "2020-06-14T17:32:49.568766Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "-3.151433947295352"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def classify_data(data):\n",
    "    \n",
    "    label_column = data.values[:, -1]\n",
    "    unique_classes, counts_unique_classes = np.unique(label_column, return_counts=True)\n",
    "\n",
    "    index = counts_unique_classes.argmax()\n",
    "    classification = unique_classes[index]\n",
    "    \n",
    "    return classification\n",
    "classify_data(X)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Isolation Tree"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:20:31.232532Z",
     "start_time": "2020-06-14T17:20:31.227383Z"
    }
   },
   "source": [
    "<img src=\"images/tree_pseudo.png\" alt=\"Drawing\" style=\"width: 400px;\"/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:50.583184Z",
     "start_time": "2020-06-14T17:32:50.575308Z"
    }
   },
   "outputs": [],
   "source": [
    "def isolation_tree(data,counter=0, max_depth=50):\n",
    "    \n",
    "    # End Loop\n",
    "    if (counter == max_depth) or data.shape[0]<=1:\n",
    "        classification = classify_data(data)\n",
    "        return classification\n",
    "    \n",
    "    else:\n",
    "        # Counter\n",
    "        counter +=1\n",
    "        \n",
    "        # Select feature\n",
    "        split_column = select_feature(data)\n",
    "        \n",
    "        # Select value\n",
    "        split_value = select_value(data,split_column)\n",
    "\n",
    "        # Split data\n",
    "        data_below, data_above = split_data(data,split_column,split_value)\n",
    "        \n",
    "        # instantiate sub-tree      \n",
    "        question = \"{} <= {}\".format(split_column, split_value)\n",
    "        sub_tree = {question: []}\n",
    "        \n",
    "        # Recursive part\n",
    "        below_answer = isolation_tree(data_below, counter,max_depth=max_depth)\n",
    "        above_answer = isolation_tree(data_above, counter,max_depth=max_depth)\n",
    "        \n",
    "        if below_answer == above_answer:\n",
    "            sub_tree = below_answer\n",
    "        else:\n",
    "            sub_tree[question].append(below_answer)\n",
    "            sub_tree[question].append(above_answer)\n",
    "        \n",
    "        return sub_tree\n",
    "    \n",
    "    \n",
    "        \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:51.018853Z",
     "start_time": "2020-06-14T17:32:51.011520Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'feat2 <= 1.9075625954665596': [-3.151433947295352, 1.9180531945595873]}\n"
     ]
    }
   ],
   "source": [
    "tree = isolation_tree(X, max_depth=1)\n",
    "pprint(tree)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Forest\n",
    "\n",
    " 1. For each tree\n",
    " 2. Sample Data\n",
    " 3. Fit tree"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<img src=\"images/forest_pseudo.png\" alt=\"Drawing\" style=\"width: 400px;\"/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:52.624210Z",
     "start_time": "2020-06-14T17:32:52.619131Z"
    }
   },
   "outputs": [],
   "source": [
    "def isolation_forest(df,n_trees=5, max_depth=5, subspace=256):\n",
    "    forest = []\n",
    "\n",
    "    for i in range(n_trees):\n",
    "        # Sample the subspace\n",
    "        if subspace<=1:\n",
    "            df = df.sample(frac=subspace)\n",
    "        else:\n",
    "            df = df.sample(subspace)\n",
    "        \n",
    "\n",
    "        # Fit tree\n",
    "        tree = isolation_tree(df,max_depth=max_depth)\n",
    "        \n",
    "        # Save tree to forest\n",
    "        forest.append(tree)\n",
    "    \n",
    "    return forest\n",
    "        \n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:32:52.790937Z",
     "start_time": "2020-06-14T17:32:52.773512Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'feat1 <= -1.0934641413547395': [-2.5682668408939806, -2.570872933200529]},\n",
       " {'feat1 <= 1.1837443869387005': [-2.570872933200529, -1.5468790931867786]},\n",
       " {'feat2 <= -0.6646354195494713': [-2.570872933200529, -0.65953384368241]},\n",
       " {'feat2 <= 0.47356257455463835': [-2.570872933200529, 0.4738735906832869]},\n",
       " {'feat2 <= -2.5411475181387355': [-2.570872933200529, -2.3261847932986437]}]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "isolation_forest(X,n_trees=5,max_depth=1)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Path Length"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-05-06T16:51:47.691834Z",
     "start_time": "2020-05-06T16:51:47.687093Z"
    }
   },
   "source": [
    "<img src=\"images/path_pseudo.png\" alt=\"Drawing\" style=\"width: 400px;\"/>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:34:13.822150Z",
     "start_time": "2020-06-14T17:34:13.816443Z"
    }
   },
   "outputs": [],
   "source": [
    "def pathLength(example,iTree,path=0,trace=False):\n",
    "    \n",
    "    path=path+1\n",
    "    question = list(iTree.keys())[0]\n",
    "    \n",
    "    feature_name, comparison_operator, value = question.split()\n",
    "    \n",
    "    \n",
    "    # ask question\n",
    "    if example[feature_name].values <= float(value):\n",
    "        answer = iTree[question][0]\n",
    "    else:\n",
    "        answer = iTree[question][1]\n",
    "        \n",
    "        \n",
    "    # base case\n",
    "    if not isinstance(answer, dict):\n",
    "        return path\n",
    "    \n",
    "    # recursive part\n",
    "    else:\n",
    "        residual_tree = answer\n",
    "        return pathLength(example, residual_tree,path=path)\n",
    "\n",
    "    return path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:34:51.747878Z",
     "start_time": "2020-06-14T17:34:51.736654Z"
    }
   },
   "outputs": [],
   "source": [
    "tree = isolation_tree(X.head(50),max_depth=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:34:55.867902Z",
     "start_time": "2020-06-14T17:34:55.862373Z"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ins = X.sample(1)\n",
    "pathLength(ins,tree)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plotting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:36:20.969994Z",
     "start_time": "2020-06-14T17:36:20.962170Z"
    }
   },
   "outputs": [],
   "source": [
    "def makeline(data,example,iTree,path=0,line_width=1):\n",
    "    \n",
    "    #line_width = line_width +2\n",
    "    path=path+1\n",
    "    question = list(iTree.keys())[0]\n",
    "    feature_name, comparison_operator, value = question.split()\n",
    "    print(question)\n",
    "    \n",
    "    # ask question\n",
    "    if example[feature_name].values <= float(value):\n",
    "        answer = iTree[question][0]\n",
    "        data = data[data[feature_name] <= float(value)]\n",
    "    else:\n",
    "        answer = iTree[question][1]\n",
    "        data = data[data[feature_name] > float(value)]\n",
    "        \n",
    "\n",
    "    if feature_name == 'feat1':\n",
    "        plt.hlines(float(value),xmin=data.feat1.min(),xmax=data.feat1.max(),linewidths=line_width)\n",
    "    else:\n",
    "        plt.vlines(float(value),ymin=data.feat2.min(),ymax=data.feat2.max(),linewidths=line_width)\n",
    "             \n",
    "        \n",
    "    # base case\n",
    "    if not isinstance(answer, dict):\n",
    "        return path\n",
    "    \n",
    "    # recursive part\n",
    "    else:\n",
    "        if feature_name == 'feat1':\n",
    "            plt.hlines(float(value),xmin=data.feat1.min(),xmax=data.feat1.max(),linewidths=line_width)\n",
    "        else:\n",
    "            plt.vlines(float(value),ymin=data.feat2.min(),ymax=data.feat2.max(),linewidths=line_width)\n",
    "        residual_tree = answer\n",
    "        return makeline(data,example, residual_tree,path=path,line_width=line_width)\n",
    "    \n",
    "    return path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:36:21.575823Z",
     "start_time": "2020-06-14T17:36:21.570747Z"
    },
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "def make_plot(data,example,iTree):\n",
    "    plt.figure()\n",
    "    plt.plot(data['feat1'],data['feat2'],'bo',alpha=0.2)\n",
    "    plt.xlabel('feat1')\n",
    "    plt.ylabel('feat2')\n",
    "    #plt.xlim(data.feat1.min(),data.feat1.max())\n",
    "    plt.ylim(-3,3)\n",
    "    plt.xlim(-3,3)\n",
    "      \n",
    "    \n",
    "    # Plot H,v line\n",
    "    makeline(data,example,tree)\n",
    "    \n",
    "\n",
    "    # Plot the point we are looking for\n",
    "    plt.scatter(x=example.feat1,y=ins.feat2,c='r',marker='o')\n",
    "    \n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:38:40.017429Z",
     "start_time": "2020-06-14T17:38:39.710374Z"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "        feat1     feat2\n",
      "142  0.420156 -0.183191\n",
      "feat1 <= -0.4474945765116678\n",
      "feat2 <= -0.07954538642763231\n",
      "feat2 <= -0.2843688278806902\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data_plot = X.sample(10)\n",
    "ins = data_plot.sample(1)\n",
    "print(ins)\n",
    "\n",
    "tree = isolation_tree(data_plot,max_depth=50)\n",
    "make_plot(data_plot,ins,tree)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Evaluate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:38:58.128837Z",
     "start_time": "2020-06-14T17:38:53.364471Z"
    }
   },
   "outputs": [],
   "source": [
    "iForest = isolation_forest(X,n_trees=20, max_depth=100, subspace=256)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:38:59.224652Z",
     "start_time": "2020-06-14T17:38:59.221230Z"
    }
   },
   "outputs": [],
   "source": [
    "def evaluate_instance(instance,forest):\n",
    "    paths = []\n",
    "    for tree in forest:\n",
    "        paths.append(pathLength(instance,tree))\n",
    "    return paths "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:39:07.326567Z",
     "start_time": "2020-06-14T17:39:07.321069Z"
    }
   },
   "outputs": [],
   "source": [
    "def c_factor(n) :\n",
    "    \"\"\"\n",
    "    Average path length of unsuccesful search in a binary search tree given n points\n",
    "    \n",
    "    Parameters\n",
    "    ----------\n",
    "    n : int\n",
    "        Number of data points for the BST.\n",
    "    Returns\n",
    "    -------\n",
    "    float\n",
    "        Average path length of unsuccesful search in a BST\n",
    "        \n",
    "    \"\"\"\n",
    "    return 2.0*(np.log(n-1)+0.5772156649) - (2.0*(n-1.)/(n*1.0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\\begin{equation*}\n",
    "s_{(x,n)} = 2 ^ {-\\frac{E(h(x))}{c(n)}}\n",
    "\\end{equation*}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:39:08.145542Z",
     "start_time": "2020-06-14T17:39:08.140995Z"
    }
   },
   "outputs": [],
   "source": [
    "def anomaly_score(data_point,forest,n):\n",
    "    '''\n",
    "    Anomaly Score\n",
    "    \n",
    "    Returns\n",
    "    -------\n",
    "    0.5 -- sample does not have any distinct anomaly\n",
    "    0 -- Normal Instance\n",
    "    1 -- An anomaly\n",
    "    '''\n",
    "    # Mean depth for an instance\n",
    "    E = np.mean(evaluate_instance(data_point,forest))\n",
    "    \n",
    "    c = c_factor(n)\n",
    "    \n",
    "    return 2**-(E/c)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate the whole dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:39:13.360961Z",
     "start_time": "2020-06-14T17:39:10.085594Z"
    }
   },
   "outputs": [],
   "source": [
    "an= []\n",
    "for i in range(X.shape[0]):\n",
    "    an.append(anomaly_score(X.iloc[[i]],iForest,256))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate one instance with all the trees"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:39:16.106545Z",
     "start_time": "2020-06-14T17:39:16.101702Z"
    }
   },
   "outputs": [],
   "source": [
    "def instance_depth_plot(instance,outlier,forest):\n",
    "    bars1 = evaluate_instance(outlier,forest)\n",
    "\n",
    "    bars2 = evaluate_instance(instance,forest)\n",
    "\n",
    "    # width of the bars\n",
    "    barWidth = 0.3\n",
    "\n",
    "    # The x position of bars\n",
    "    r1 = np.arange(len(bars1))\n",
    "    r2 = [x + barWidth for x in r1]\n",
    "    \n",
    "    # Create cyan bars\n",
    "    plt.bar(r2, bars2, width = barWidth, capsize=7, label='Normal Sample')\n",
    "\n",
    "    # Create blue bars\n",
    "    plt.bar(r1, bars1, width = barWidth,  capsize=7, label='Outlier')\n",
    "    #sns.barplot(x=r1, y=bars1,capsize=7, label='Outlier')\n",
    "\n",
    "    \n",
    "    #sns.barplot(x=r2, y=bars2, label='Normal')\n",
    "    \n",
    "    # general layout\n",
    "\n",
    "    plt.ylabel('Tree Depth')\n",
    "    plt.xlabel('Trees')\n",
    "    plt.legend()\n",
    "\n",
    "    # Show graphic\n",
    "    plt.savefig('images/normal_vs_outlier.png')\n",
    "\n",
    "    plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:39:33.945864Z",
     "start_time": "2020-06-14T17:39:33.373326Z"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABCUAAAIACAYAAAC1udYtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nOzde5hWZb038O8AIgOCB5CDgoQwtMO2lqJpvejOQ4JIaprt8GyZsU0UEdQOYipoeIgyT9u27lLcYZh5StMsTd3Z637LPCajYYogJ0MSBmSGef9wO0UwOAMzs+aZ+Xyui+uaZ611r/Vb677XPM98WWs9ZcuWLasNAAAAQAvrUHQBAAAAQPsklAAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAKIZQoAZWVlUWXQBPSn22L/mxb9Gfbo0/bFv3ZtujPtkV/ti0t2Z9CCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQnYouoKVVV1dnxYoVRZfRKF26dMlbb71VdBn8r27duqVTp3Z36gAAADS5dvWXVXV1df76179mm222SVlZWdHlNNiWW26ZLl26FF0GSWpra7Ns2bJ0795dMAEAALCZ2tXtGytWrCi5QILWpaysLNtss03JXW0DAADQGrWrUCKJQILNZgwBAAA0jXYXSgAAAACtg1ACAAAAKIRQogQMHz48Bx54YGpra9eZ/uijj2abbbZJdXV1QZX9zcyZMzNs2LB65y9ZsiTjx4/P0KFD07t37+y55565/PLLs2bNmhas8m+22WabPPzww4VsGwAAgHf5+oAkc5dXZ96KmhbbXv9uHTOoR+MO/f/8z//kBz/4QU488cTmKaqZHX300enXr19mzZqV7bbbLk899VQmT56cpUuX5pJLLim6PAAAAAoglEgyb0VNxty/pMW2d/fIXo0OJQYMGJBvfvObGTNmTHr27NlMlTWP5557Lr/73e/y8ssv19U+cODALF++PJMnT87UqVPToYOLdgAAANobfwmWiNNOOy1bbbVVzj///HqXWbZsWc4444xUVFRkwIABOeWUU/KXv/wlybu3egwbNiyTJk3KTjvtlEsuuSTjxo3L1772tZx88snp169f9t577zzzzDO56KKLstNOO2WXXXbJXXfdVbf+//t//29GjRqVfv36ZYcddsiRRx6Z+fPnv2/t731bxa9+9at1ph955JH59a9/XTd/zpw5Oeqoo9K/f//06dMnI0eOzAsvvLBO/bfcckuGDh2agQMH5qqrrsqjjz6aPffcM/3798+//du/Ze3atUmS0aNHZ9q0aRk1alT69u2bkSNH5o9//OMG61u9enXOPffcDB48OIMGDcqJJ56YRYsWve9+AQAAsHmEEiWia9euufTSS3PrrbfmiSee2OAyxxxzTJ555pn86Ec/yp133pnKysqceuqpdfPnz5+f5cuX55FHHsnYsWOTJDfccEP23nvvPPbYY+nRo0cOPfTQvPnmm/nFL36RT37ykznzzDNTW1ubv/71rzn66KOz33775Te/+U1+8pOf5JVXXskVV1zxvrUPGzYs++23X774xS/mE5/4RM4///z88pe/zBZbbJGKioqUlZWltrY2n//85zNgwIA8+uij+fnPf56ampp1QphFixblrrvuyj333JMJEybkggsuyDe+8Y1ce+21ufbaa3Pbbbfl5z//ed3y3/nOd3LooYfmkUceyQ477JCjjjoqq1atWq++Cy+8ME8++WRmzZqVe++9N2vXrs3nPve59Z7hAQAAQNMSSpSQ0aNH51Of+lQmTpy43sMtn3322Tz++OO59tprs8cee2T33XfPv//7v+eBBx6ou9ogSc4444wMGjQoAwcOTJJ8+MMfzpe+9KUMHjw4Rx55ZKqqqnLJJZdk6NChOeWUU/Lmm29m6dKlWblyZSZOnJhzzjknH/jAB7L33nvn05/+9Drr3pjbbrstU6ZMSU1NTb773e/mM5/5TP75n/85v/zlL5MkK1euzAknnJCLLroogwYNykc+8pGMHTt2nfVXV1fnwgsvzNChQ/OFL3whNTU1+eIXv5jhw4dnzJgx+eAHP5jKysq65ffff/+cdtpp+eAHP5jvfOc7WbZsWX7xi1+sU9fKlStzww035Morr8zw4cMzbNiwXH/99fnjH/+Y3/zmN43rIAAAABrFMyVKzPTp07P33nvnuuuuy2677VY3fc6cOenevXs++MEP1k0bOnRottlmm8yZMyfbbbddkmSnnXZaZ30f+MAH6n4uLy9P796906VLl7rXybu3N+y4444ZO3Zsrr766jzzzDN58cUX8+yzz2b48OENqnvLLbfMhAkTMmHChLz22mt58MEHc8011+SYY47JU089lT59+uTkk0/OrFmz8vvf/z6VlZX5wx/+UFf3P9b7Xm0DBgxYp/7Vq1fXvf7Yxz5W93P37t0zePDgzJkzZ531vfLKK3nnnXdyyCGHrDN91apVeemll/Lxj3+8QfsHAABA4wklSszAgQMzceLEfOtb38qVV15ZN/29IOEfrV27NjU1NfUu16nTukPgvec7/KP58+fnk5/8ZHbdddfsv//+OeGEE/LAAw/UeyvJ37vrrruyYMGCultJBgwYkJNPPjmHHXZYhg0blieeeCIHHHBA9t9//2y77bYZPXp0jjrqqMyZMyczZszYaL0be0Bmx44d13ldU1Oz3vLvXXFy7733pkePHuvMK7UHigIAAJQat2+UoPHjx6dv37656KKL6qZVVFTkr3/9a1588cW6aX/84x+zfPnyDBkyZLO3ec8996R79+758Y9/nHHjxuXjH/94XnnllQY9d+HVV1/N9OnT8/bbb68zvVu3bunUqVN69uyZxx57LK+//nruueeejB8/Pv/yL/+SefPmbdZzHZ555pm6n996663MnTs3u+yyyzrLDBo0KB07dszSpUuz8847Z+edd07Pnj3z1a9+Na+99tombxsAAID350qJEtS5c+dcfvnlOeyww+qmVVRU5OCDD864ceNy+eWXJ0kmTpyYffbZJ7vuumseffTRzdrmdtttlwULFuRXv/pVBg0alJ/+9Ke56667suuuu75v22OOOSbXXXddDj/88Jx33nmpqKjIq6++mmuuuSbDhg3LJz7xiTz55JNZuXJl7r777uyxxx555JFHcsMNN9TdprEp7rjjjuy3334ZPnx4pk6dmh122CH/8i//ss4y3bt3zwknnJBJkybl29/+dvr165cLLrggzz33XAYPHrzJ226N5i6vzrwVNe+/4D/o361jo7/CFgCgPfJ5i8YwXt7VdvZkM/Tv1jF3j+zVotvbXPvtt1+OOuqozJ49u27addddl8mTJ+ewww5Lhw4dcsghh+SSSy7Z7G0lyRFHHJH//u//zoknnpgk2X333TNt2rRcdNFFqaqq2mjbbbfdNj//+c8zbdq0nH766Vm8eHF69uyZQw89NNdcc03Kysqy11575ZxzzsnkyZOzevXqDBs2LJdffnlOO+20Tb5i4aijjsrNN9+cs88+Ox//+Mdz++23Z4sttlhvuYsvvjjnn39+Tj755KxevTp77bVX7rjjjs0KRFqjeStqMub+JY1ud/fIXm3qlx4AQHPxeYvGMF7eVbZs2bJ2872Hb731Vrbeeuuiy2i0VatW1fvMCDZs9OjR2WefffL1r3+9Wda/OWOpsrIyFRUVTVzR+3t0wepN/qU3ot+WzVBR21BUf9I89Gfbo0/bFv3ZtrTF/mzPn7faYn82t9Y8XlqyPz1TAgAAACiEUAIAAAAoRNu5EQX+zr333lt0CQAAALwPV0oAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEqUkMWLF2fixIkZNmxY+vbtm+HDh+fSSy9NVVVVg9rX1tbmxhtvzNq1a5Mk48aNy5e+9KUkySWXXJKRI0c2W+0AAADwjzoVXUBrULZofsqWLmyx7dX27JPa3js0qs38+fNz8MEHZ+DAgbn++uuz00475bnnnsvFF1+cBx54IPfcc0+6du260XU8/vjjOeuss3L88cenQ4d186jTTz89X/7ylxu9LwAAALCphBJJypYuTNdLJ7TY9lae++1GhxKTJk1K//7989Of/jSdOr3bbQMHDszHP/7x7LPPPrnssssyZcqUja6jtra23nlbbbVVo+oBAACAzeX2jRKwePHi3HfffTnzzDPrAon3bLPNNhk3blxuvvnmPPLII9lmm21SXV1dN/+9WzT+/Oc/Z8yYMUmSXr165dFHH11nPf94+8ZvfvOb7L///unbt2/23nvv/OhHP1pnnV/+8pczYsSI7LzzznnuueeaY7cBAABo44QSJeDpp5/O2rVrs8cee2xw/j777JMlS5bklVdeqXcd/fv3zw9/+MMkyfPPP5+Pfexj9S67cOHCHH300fnsZz+b//7v/87kyZMzefLk3HfffXXL3HbbbZk8eXJmz56dD33oQ5u2YwAAALRrbt8oAcuWLUvy7lURG/Le9KVLl9a7jo4dO2bbbbdNkvTu3Xu9Ky7+3g033JD/83/+T8aNG5ck2XnnnTNnzpxce+21GTVqVJJk1113rbvyAgAAADaFUKIEvBc6LFy4MDvuuON68xcsWJAk2W677Zpke3PmzMkvfvGLdbZVXV2dXr161b3eaaedmmRbAAAAtF9CiRLwkY98JB07dszvf//7DYYSv//979OzZ88MGTJkvXk1NTWN3l51dXWOPPLITJ48eZ3pf/+NHV26dGn0egEAAODveaZECejZs2cOP/zwfOtb31rnIZZJ8tZbb+V73/tejj322HTu3DlJ8vbbb9fN//vnTJSVlTVoexUVFfnTn/6UnXfeue7fQw89lJtvvnnzdwYAAAD+l1CiREybNi0rVqzIEUcckcceeyyvvfZaHnzwwYwePTr9+/fPOeeck3/6p39KeXl5rrzyyrzyyiv53ve+l6effrpuHe997ecf/vCHrFq1qt5tfeELX8jTTz+db37zm3n55Zdz5513ZsqUKdlhh8Z9jSkAAABsjNs3ktT27JOV5367RbfXWH369MmDDz6YK664IuPGjcvixYvTv3//HHXUUTnjjDNSXl6eJPnOd76TCy+8MDfccEMOPfTQfPnLX878+fOTJMOGDcsBBxyQUaNG5fvf/36929ppp50ya9asXHDBBbn66qvTu3fvnHfeefnCF76waTsMAAAAGyCUSFLbe4fU9m79VwH07Nkz06ZNy7Rp0+pd5uijj87RRx+9wXmdO3fO7bffXvf605/+dN3P5513Xs4777y61/vtt19+9atfbXA91157bWNLh80yd3l15q1o/PNR+nfrmEE9/JoDAIDWyqd1oNWbt6ImY+5f0uh2d4/sJZQAAIBWzDMlAAAAgEIIJQAAAIBCCCUAAACAQgglAAAAgEK0u1Citra26BIoccYQAABA02hXoUS3bt2ybNkyf1SyyWpra7Ns2bJ069at6FIAAABKXot/V15tbW0uvPDCDB48OMcee2zOPffcvPbaa3Xz58+fn9133z1XXHHFem0/9alPZfvtt697fdxxx2XkyJEN3nanTp3SvXv3LF++fPN2ooUtX748PXr0KLoM/lf37t3TqZOvmQQAANhcLfqX1dy5czN9+vQ8++yzGTx4cJLk0ksvrZv//PPP59xzz83kyZPXa/vnP/853bt3z8yZMzerhk6dOmXrrbferHW0tEWLFmXAgAFFlwEAAABNqkVDidmzZ2fMmDHp27fvevPWrFmTb37zm5kwYUL69Omz3vynn346HTt2zLhx4/LWW29l//33z0knnZSOHTu2ROkAAABAE2vRUGLSpElJkieffHK9eXfeeWd69eqVT37ykxtsW1NTk7322ivjx4/P6tWrM2HChHTr1i2f//znm7VmAAAAoHm0mhvj/+u//itf/epX651/+OGH1/3cuXPnjB07NrNmzao3lKisrGzyGovU1vanvSuiP6s6bv/+C22oXVVVKitfbeJqGllDK6/d+dm26M+2R5+2LfqzbWlr/dnaP7M0t7bWn82ttY+XpurPioqKjc5vFaHEiy++mJqamuy+++71LvOzn/0sFRUVdTtUW1u70YcNvt+Ol5LKyso2tT/tXVH9+caC1UlWNLpdeXl5KvoVO/5ac+3Oz7ZFf7Y9+rRt0Z9tS1vsz9b8maW5tcX+bG6teby0ZH+2iq8E/d3vfpfhw4enrKys3mVefvnlXH/99ampqcmqVavy4x//OAcddFALVgkAAAA0pVYRSrz22mvp16/fetOvv/76XH/99UmSU045JVtvvXXGjh2bY445JrvuumsOO+ywli4VAAAAaCKF3L4xZcqUdV5v6CtAk+TUU0+t+7lLly75xje+0ax1AQAAAC2nVVwpAQAAALQ/QgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQnYouAACaytzl1Zm3oqbR7fp365hBPbwlAgC0NJ/AAGgz5q2oyZj7lzS63d0jewklAAAK4PYNAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEJ2KLgCA1mfu8urMW1HT6Hb9u3XMoB7eWgAAaBifHAFYz7wVNRlz/5JGt7t7ZC+hBAAADeb2DQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBCdWnqDtbW1ufDCCzN48OAce+yxSZJPfepT2X777euWOe644zJy5Mj12t5000352c9+lpqamowcOTKnnHJKysrKWqx2AAAAoOm0aCgxd+7cTJ8+Pc8++2wGDx6cJPnzn/+c7t27Z+bMmRtt+/jjj+ehhx7KD3/4w3To0CHjx4/PoEGDctBBB7VE6QAAAEATa9HbN2bPnp0xY8bkwAMPrJv29NNPp2PHjhk3blzGjh2b73//+6mpqVmv7cMPP5yDDz445eXl2XLLLTNmzJjcf//9LVk+AAAA0IRa9EqJSZMmJUmefPLJumk1NTXZa6+9Mn78+KxevToTJkxIt27d8vnPf36dtgsXLsyee+5Z97p3795ZtGhRyxQOAAAANLkWf6bEPzr88MPrfu7cuXPGjh2bWbNmrRdK1NbWrte2Q4f6L/SorKxsuiJbgba2P+1dEf1Z1XH7919oQ+2qqlJZ+WoTV9PIGlp57W3x/Gztx7ze7TdB3W2xP9s7fdq26M+2pa31Z6m+fzaVIvpzZZftsnBNx0a367NFTbquerMZKmq41j5emqo/KyoqNjq/8FDiZz/7WSoqKuoKra2tTadO65fVp0+fLFmypO71okWL0rt373rX+347XkoqKyvb1P60d0X15xsLVidZ0eh25eXlqehX7PhrzbW31fOzNR/zjdncuttqf7Zn+rRt0Z9tS1vsz1J9/2wKRfXnowtW5+hfLHn/Bf/B3SN7ZbeKns1QUcO15vHSkv1Z+FeCvvzyy7n++utTU1OTVatW5cc//vEGH16533775f77709VVVXeeeed3Hvvvdlvv/0KqBgAAABoCoVfKXHKKafksssuy9ixY1NdXZ0DDjgghx12WJLk9ttvzwsvvJCvf/3rGTFiRF566aWceOKJqa6uzr777pvRo0cXXD0AAACwqQoJJaZMmVL3c5cuXfKNb3xjg8sdeeSR67w+6aSTctJJJzVrbQAAAEDLKPz2DQAAAKB9EkoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIXoVHQBAAAAf2/u8urMW1HT6Hb9u3XMoB7+xIFS4owFAABalXkrajLm/iWNbnf3yF5CCSgxbt8AAAAACiGUAAAAAAohlAAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAK0anoAgAAaLi5y6szb0VNo9v179Yxg3r46NfebMp4MVbaJ79bKIrRAwBQQuatqMmY+5c0ut3dI3v5w6Ed2pTxYqy0T363UBS3bwAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIVo0e9uqa2tzYUXXpjBgwfn2GOPzapVq3LZZZfl+eefz9q1a/PhD384kyZNSpcuXdZre/zxx2f16tXp1OndkkeOHJnjjjuuJcsHAAAAmlCLhRJz587N9OnT8+yzz2bw4MFJkptuuinV1dWZOXNmamtrM2XKlPzgBz/Iqaeeuk7bqqqqzJs3Lw888EBdKAEAAACUthb7C3/27NkZM2ZM+vbtWzftox/9aPr165cOHd69i2To0KH505/+tF7b5557Ll27ds2ECROyZMmS7Lnnnvm3f/u3DV5RAQAAAJSGFnumxKRJk3LIIYesM23vvffOwIEDkyQLFizIj370oxxwwAHrtV25cmX22GOPXHLJJfnP//zPLFy4MNdcc02L1A0AAAA0j1ZxL8QLL7yQyZMn57Of/WxGjBix3vx99903++67b93rE088Meecc07OOuusetdZWVnZLLUWpa3tT3tXRH9Wddx+09pVVaWy8tUmrqaRNbTy2tvi+dnaj3m922+Cuttif7Z3ba1PS/X8bCptrT+b26aMl5YcK/X1Z6mO81KtOynd99D2fsybU1P1Z0VFxUbnFx5KPPDAA5k+fXrOPvvsjBw5coPLPProo+nWrVt23333JO8+MPP9ni3xfjteSiorK9vU/rR3RfXnGwtWJ1nR6Hbl5eWp6Ffs+GvNtbfV87M1H/ON2dy622p/tmdtsU9L9fxsCm2xP5vbpoyXlhorG+vPUh3npVp3Urrvoe35mDenluzPQr8S9KGHHsoVV1yR7373u/UGEkmycOHCfPe7382qVatSU1OTW2+9NQceeGALVgoAAAA0tUKvlLjmmmtSW1ubqVOn1k3bbbfdMnny5Nx+++154YUX8vWvfz2f+cxnMn/+/Bx//PGpqanJHnvskS9+8YsFVg4AAABsrhYPJaZMmVL38+23317vckceeWTdzx06dMj48eMzfvz4Zq0NAAAAaDmF3r4BAAAAtF9CCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBCdii6AzTN3eXXmrahpdLv+3TpmUA/dDwAAQHH8VVri5q2oyZj7lzS63d0jewklAAAAKJTbNwAAAIBCCCUAAACAQgglAAAAgEIIJQAAAIBCCCUAAACAQgglAAAAgEIIJQAAAIBCCCUAAACAQgglAAAAgEIIJQAAAIBCCCUAAACAQgglAAAAgEJ0auiCL7/8cl566aW88847qa2tXWfepz/96SYvDAAAAGjbGhRK3HLLLbnqqqvSvXv3dOvWbZ15ZWVlQgkAAACg0RocSkyYMCH/+q//2tz1AAAAAO1Eg54p8c4772TEiBHNXQsAAADQjjQolBg9enRmzZrV3LUAAAAA7Ui9t2+ccsopKSsrS5JUV1fnueeey0MPPZR+/fqlQ4d1s4x///d/b94qAQAAgDan3lBir732Wuf1Pvvs0+zFAAAAAO3HRq+UeM+9996bgw46KJ07d15nmaqqqtx1113NVx0AAADQZtUbSrz55pupqqpKklx00UUZNGhQtt5663WWmTNnTr73ve/lc5/7XPNWCQAAALQ59YYSTz31VM4777y650qcdNJJG1zu0EMPbZ7KAAAAgDat3lBi//33z5133pm1a9fmiCOOyE033ZRtt922bn5ZWVnKy8vXu3oCAAAAoCHqDSWSpG/fvkmS3/72t3XTli5dmg4dOqwTUAAAAAA01kZDiffU1NTkuuuuy5133pnly5cnSbbbbrt89rOfrfe2DgAAAICNaVAoceWVV+aRRx7J+PHj86EPfSg1NTV5/vnn8/3vfz9r1qzJl770peauEwAAAGhjGhRK3H///bnsssuy++67100bOnRodthhh5x//vlCCQAAAKDROjRkofLy8nTqtH5+0b1797pv5wAAAABojAaFEl/5yldy8cUX59e//nX+8pe/ZPny5fmf//mfXHzxxfnXf/3XvP7663X/AAAAABqiQbdvTJkyJUkyadKkuisjamtrkyQvvfRSrr322tTW1qasrCxPPPFEM5UKAAAAtCUNCiV++tOfNncdAAAAQDvToFCiX79+SZI//OEPefXVV7P//vvnjTfeyIABA9K5c+dmLRAAAABomxoUSrz55ps566yz8qc//Slr1qzJ7rvvnmuvvTYvvfRSrrrqqgwYMKC56wQAAADamAY96PLyyy9Pv3798sADD2TLLbdMklxwwQUZMmRIrrjiimYtEAAAAGibGhRKPPnkkznllFPSpUuXumlbbbVVvvKVr+QPf/hDsxUHAAAAtF0NCiU6dOiQVatWrTd9yZIldVdOAAAAADRGg0KJgw8+OJdffnnmzJmTsrKyrFixIr/97W9zySWX5IADDmjuGgEAAIA2qEEPujz99NNz9dVX5+STT86aNWty3HHHpWPHjjnssMNy+umnN3eNAAAAQBvUoFBiiy22yJlnnpkvf/nLef3111NTU5P+/funa9euzV0fAAAA0Ea9byixZs2aPPXUU5k7d25WrlyZbt26ZfDgwes89BIAAACgsQkfsK4AACAASURBVDYaStx777256qqr8pe//CVdu3ZNt27dsnLlyqxYsSK9evXK+PHjc/DBB7dUrQAAAEAbUm8ocf/992fq1Kk54YQTcvjhh6dPnz518xYsWJC77rorF198cbbaaqt84hOfaPAGa2trc+GFF2bw4ME59thjU1NTkxkzZuSJJ55ITU1NjjnmmBx55JHrtWvocgAAAEBpqDeUuPXWWzNu3Lgcd9xx683r169fTj311HTr1i233HJLg0OJuXPnZvr06Xn22WczePDgJMkdd9yR1157Lf/1X/+VlStX5gtf+EL+6Z/+Kbvssss6bRu6HAAAAFAa6v1K0FdeeSX77rvvRhuPGDEiL7/8coM3Nnv27IwZMyYHHnhg3bSHH344hx56aDp16pQePXrkoIMOyn333bde24YuBwAAAJSGeq+UWL16dbbaaquNNt5qq62yfPnyBm9s0qRJSZInn3yybtrChQvXuTWkd+/eeemll9Zr29Dl3lNZWdngukpBfftT1XH7TVpfVVVVKitf3ZySNsvKLttl4ZqOjW7XZ4uadF31ZjNU1LKKGJ+lOlaS1l97ff1ZyuO8tR/zerffBHUX9f5RyuOltWtr52ipnp9Npa19xmtumzJeWnKstLXPuKVad1K676Ht/Zg3p6bqz4qKio3O3+iDLsvKypqkiI1Zu3btetM6dFj/Ao6GLvee99vxUlJZWVnv/ryxYHWSFY1eZ3l5eSr6FXeMHl2wOkf/Ykmj2909sld2q+jZDBW1nI31Z3Mq1bGStO7aN9afpTzOW/Mx35jNrbuo8zMp7fHSmrXFc7RUz8+mUOQ5Wqo2Zby01Fhpi59xS7XupHTfQ9vzMW9OLdmfGw0lfvjDH6a8vLze+VVVVZtdQN++fbNkyd8+ECxevDi9e/fe5OUAAACA0lBvKPHRj340L7744vuu4KMf/ehmFbDvvvvm7rvvzogRI1JVVZUHH3ww55577iYvBwAAAJSGekOJ6667rkUKOPLII/P666/nmGOOSXV1dY444ojsvvvuSZLrr78+SXLqqadudDkAAACg9Gz09o3mMmXKlL8V0KlTzjrrrA0ud+qppzZoOQAAAKD01P+kSAAAAIBmJJQAAAAACiGUAAAAAArR4GdKLF26NHfeeWdeffXVjB8/Pv/v//2/fOADH/Bd0QAAAMAmadCVEs8//3yOOuqoPPnkk3nwwQdTVVWV3/3udzn55JPzxBNPNHeNAAAAQBvUoFBixowZOeGEE3Lttddmiy22SJKcc845Of7443P11Vc3a4EAAABA29SgUGLOnDk58MAD15t+yCGH5JVXXmnqmgAAAIB2oEGhxLbbbpu5c+euN/2pp57K9ttv3+RFAQAAAG1fgx50efzxx2fatGk54YQTsnbt2vz2t7/NG2+8kdtuuy2nnXZac9cIAAAAtEENCiWOOOKI9OrVKzfffHO6dOmSq6++OgMHDszXvva1HHTQQc1dIwAAANAGNfgrQUeMGJERI0Y0Zy0AAABAO9KgZ0okyWOPPZavfOUrOeyww7JgwYJcd911ueOOO5qzNgAAAKANa1Aocd999+WCCy7IRz7ykbz55ptZu3ZtevbsmRkzZuTWW29t7hoBAACANqhBocQPf/jDnHvuufniF7+Yjh07Jkk++9nP5vzzz8+sWbOatUAAAACgbWpQKDFv3rx86EMfWm/60KFD8+abbzZ5UQAAAEDb16BQYsiQIXn88cfXm3733XdnyJAhTV4UAAAA0PY16Ns3zjjjjEyYMCFPPvlk1qxZk//4j//Ia6+9lhdffDFXXnllc9cIAAAAtEENulLiIx/5SGbPnp0hQ4ZkxIgR+etf/5rddtstt912W4YPH97cNQIAAABtUIOulDj99NNz1lln5dRTT23uegAAaKPmLq/OvBU1jWrTv1vHDOrRoI+sAJSgBv2GnzNnTjp18mYAAMCmm7eiJmPuX9KoNneP7CWUAGjDGvQb/jOf+UzOPffcHHHEEenXr186d+68zvw999yzWYoDAAAA2q4GhRI33nhjkuSyyy5bb15ZWVmeeOKJpq0KAAAAaPPqDSXuvffeHHTQQencuXN++9vftmRNAAAAQDtQ77dvXHTRRXn77bdbshYAAACgHak3lKitrW3JOgAAAIB2pt5QInn3eREAAAAAzWGjD7ocNWpUg1biQZcAAABAY200lJg6dWq6d+/eUrUAAAAA7Ui9oURZWVk++tGPZrvttmvJegAAAIB2woMuAQAAgELUG0qMHj06W265ZUvWAgAAALQj9d6+cf7557dkHQAAAEA7s9GvBAUAAABoLkIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQQgkAAACgEEIJAAAAoBBCCQAAAKAQnYouAIpStmh+ypYubHS72p59Utt7h2aoCKD0zF1enXkrahrdrn+3jhnUw8cQSoNxDtB8/Jak3SpbujBdL53Q6HYrz/22UALgf81bUZMx9y9pdLu7R/byxxolwzgHaD5u3wAAAAAKIZQAAAAACiGUAAAAAAohlAAAAAAKIZQAAAAACiGUAAAAAApR+HcU3Xvvvbn11lvrXr/99ttZtGhR7rnnnvTs2bNu+owZM/LQQw+lR48eSZKBAwdm2rRpLV4vAAAA0DQKDyVGjx6d0aNHJ0mqq6vzpS99KSeccMI6gUSSPP3005k6dWp23XXXIsoEAAAAmljhocTf+8EPfpDtttsun/nMZ9aZ/s4772TOnDm55ZZbMm/evAwYMCATJkxI3759C6oUAAAA2Fyt5pkSy5Yty6233poJEyasN2/x4sUZPnx4TjvttMycOTMf/vCHc/bZZ6e2traASgEAAICm0GqulLjjjjuy7777Zscdd1xv3o477pgZM2bUvT722GNz4403Zv78+RtcPkkqKyubrdYi1Lc/VR2336T1VVVVpbLy1c0pabO0hrp3qqpK102s4dXNHF9FjM/WcMw3VWuvva2dn0np1t4UdRf1/tGej3lza2vnaKnWnWxa7f9Yt/fQRtbQBMe8OTk//7ddiY6VpPj30PZ+zJtTU/VnRUXFRue3mlDiwQcfzNlnn73BeZWVlamsrMwhhxxSN622tjadOtVf/vvteCmprKysd3/eWLA6yYpGr7O8vDwV/Yo7Rq2h7g4vvL1J7crLyzdrfG2sP5tTazjmm6o1194Wz8+kdGvf3LqLOj+T9nvMm1tbPEdLte5k02r/+7q9hzbe5h7z5uT8/Jui605K9z20PR/z5tSS/dkqbt9Yvnx55s2bV+9DLDt06JArrrgir7/+epLk9ttvz5AhQ9KnT5+WLBMAAABoQq3iSol58+alV69e61z58Pzzz2fq1KmZOXNmBg8enLPPPjsTJ07M2rVr07t371x88cUFVgwAAABsrlYRSgwbNiw/+clP1ps2c+bMutejRo3KqFGjWro0AAAAoJm0its3AAAAgPZHKAEAAAAUQigBAAAAFEIoAQAAABRCKAEAAAAUQigBAAAAFEIoAQAAABSiU9EFAMA/Kls0P2VLFza63dDyXpG3AwCUDqEEAK1O2dKF6XrphEa363rWlUn6NH1BAAA0C/+dBAAAABRCKAEAAAAUQigBAAAAFEIoAQAAABRCKAEAAAAUQigBAAAAFEIoAQAAABRCKAEAAAAUQigBAAAAFEIoAQAAABRCKAEAAAAUQigBAAAAFEIoAQAAABRCKAEAAAAUQigBAAAAFEIoAQAAABRCKAEAAAAUQigBAAAAFEIoAQAAABRCKAEAAAAUQigBAAAAFKJT0QVAqVlVU5vfLljd6Hb9u3XMoB5OOWjryhbNT9nShY1uN7S8V/xfAQDQ3vgLCRppdU0y5v4ljW5398heQgloB8qWLkzXSyc0ul3Xs65M0qfpCwIAaMX8lwwAAABQCKEEAAAAUAihBAAAAFAIoQQAAABQCKEEAAAAUAihBAAAAFAIoQQAAABQCKEEAAAAUAihBAAAAFAIoQQAAABQCKEEAAAAUAihBAAAAFAIoQQAAABQCKEEAAAAUAihBAAAAFAIoQQAAABQCKEEAAAAUAihBAAAAFCITkUXkCQzZszIQw89lB49eiRJBg4cmGnTpq2zzLPPPpvp06dn1apV6dWrVy688ML06tWriHIBAACAJtAqQomnn346U6dOza677rrB+WvWrMm5556bqVOnZrfddsvs2bNz8cUXZ8aMGS1cKQAAANBUCr9945133smcOXNyyy23ZOzYsTnnnHPyxhtvrLPM888/n27dumW33XZLkhx22GF58skns2zZsiJKBgAAAJpA4VdKLF68OMOHD89pp52WnXbaKbfcckvOPvvs3HzzzSkrK0uSLFy4MH369Klrs8UWW2TbbbfN4sWLs80222xwvZWVlS1Sf0upb3+qOm6/SeurqqpKZeWrm1PSZmkNde9UVZWum9Bu7dqaTdre39dexPhsDce8T6qz5V//0uh2O2/dL0nnRrdrqXHems/PUj3mpXx+Fll7EVrDOH8/rfkc3aTtl2jdyabV/o91t9f30E3VFMe8OTk//7ddiY6VxGfczdHaa2+q/qyoqNjo/MJDiR133HGd2zCOPfbY3HjjjZk/f3523HHHJMnatWs32LZDh/ov9Hi/HS8llZWV9e7PGwtWJ1nR6HWWl5enol9xx6g11N3hhbc3rV2HjpvU7r3aN9afzal1HPPfp+t3vtrodu+cdWWSPu+73D9qiXHe2s/PUj3mpXx+FlV7UVrDON+Y1n6ObopSrTvZtNr/vu72/B66qTb3mDcn5+ffFF13svm1Oz8brzXX3pL9WfjtG5WVlfnZz362zrTa2tp06vS3vKRv375ZsmRJ3evq6uosW7YsvXv3brE6AQAAgKZVeCjRoUOHXHHFFXn99deTJLfffnuGDBmyzu0aH/7wh7N8+fI8/fTTSZK77ror//zP/5zu3bsXUjMAAACw+Qq/fWPw4ME5++yzM3HixKxduza9e/fOxRdfnMWLF+fMM8/MjBkzsv322+db3/pWLrvsslRVVWXrrbfOBRdcUHTpAAAAwGYoPJRIklGjRmXUqFHrTZ85c2bdz7vsskv+8z//swWrAgAAAJpT4bdvAAAAAO2TUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKESnogugtJUtmp+ypQsb3W5oea/IxCgVGxvnO1VVpcMLb29wnnFOKSnl3+eleo621WO+Ma2hdlpWqZ6fyaaN89ZQdylzzNsnoQSbpWzpwnS9dEKj23U968okfZq+IGgGGxvnXTfSzjinlJTy7/NSPUfb6jHfmNZQOy2rVM/PZNPGeWuou5Q55u2TSAkAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAACiEUAIAAAAohFACAAAAKIRQAgAAAChEp6IL4F1li+anbOnCDc7bqaoqHV54e4Pzhpb3imypfdnYWNkYYwUASpf3f2i9nJ+bRyjRSpQtXZiul07Y4LyuG2nX9awrk/RplpponTY2VjbGWAGA0uX9H1ov5+fmEcsAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFaBVfCXrffffl5ptvTllZWbp06ZKJEydm2LBh6ywza9as3HTTTenZs2eSpGvXrrnhhhuKKBcAAABoAoWHEn/+85/z3e9+NzfffHN69eqVxx9/POecc07uvvvudZZ75plncuaZZ2bkyJEFVQoAAAA0pcJDiS222CJf+9rX0qtXryTJhz70oSxdujRr1qzJFltsUbfc008/nRUrVuSWW27JtttumzPOOCNDhgwpqmwAAABgMxUeSuywww7ZYYcdkiS1tbWZMWNG9t1333UCiaqqqgwcODAnnXRSdt111zz44IM588wzc9ttt6Vr165FlQ4AAABshsJDifdUVVXlm9/8ZhYtWpTvfOc768wrLy/PVVddVff6oIMOyo033pjnn38+w4cP3+D6Kisrm7XeprZTVVU2JV5Zu7Zmk7ZXVVWVyspXN6nt3yvVupNia9+c8emYN5xxXrq1l+r5mTjmDWWcl27dScvW/o91F/EZr6rj9pvWro0c883R3sZ5a6h7UzXFOC/iPbQ1HPNSHucb01S/bysqKjY6v1WEEm+88UbOOuusDBo0KNdcc026dOmyzvwFCxbk17/+dT73uc/VTautrU2nTvWX/3473tp0eOHtTWvXoeMmtSsvL09Fv80/RqVad1Jc7ZWVlZs1Ph3zhjPOS7f2Uj0/E8e8oYzz0q07adna/77upjhHN8UbC1YnWdHodm3hmG+u9jbOW0Pdm2pzx3lR76Gt4ZiX8jivT0v+vi38K0HfeuutnHrqqfnkJz+ZqVOnrhdIJO8e9Ouuuy7PPfdckuTxxx/PqlWrsssuu7R0uQAAAEATKfxKidtvvz0LFy7Mww8/nIcffrhu+re//e1MmDAhM2bMyPbbb59p06blkksuyZo1a9KtW7dMnz59nedOAAAAAKWl8FDi5JNPzsknn7zBeTNnzqz7eZ999sk+++zTUmUBAAAAzazw2zcAAACA9kkoAQAAABRCKAEAAAAUQigBAAAAFEIoAQAAABRCKAEAAAAUQigBAAAAFKJT0QUAAEBLKFs0P2VLFza63dDyXvF/eZQK45xSI5QAAKBdKFu6MF0vndDodl3PujJJn6YvCJqBcU6pEYUBAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIUQSgAAAACFEEoAAAAAhRBKAAAAAIXoVHQB73nsscdyzTXX5J133smQIUPy9a9/PVtttVWjlwEAAABKQ6u4UuIvf/lLLrroolx66aWZPXt2dtxxx1x99dWNXgYAAAAoHWXLli2rLbqI+++/Pz//+c/z7W9/O0kyf/78HHPMMfnlL3+ZsrKyBi8DAAAAlI5WcaXEwoUL07t377rXvXv3zooVK7JixYpGLQMAAACUjlYRSqxdu3aD0zt27NioZQAAAIDS0SpCib59+2bp0qV1rxcvXpwePXqkvLy8UcsAAAAApaNVhBIf+9jH8uyzz+bVV19NkvzkJz/Jvvvu2+hlAAAAgNLRKh50mSSPP/54rr766lRXV2fHHXfMBRdckNdffz1Tp07NzJkz611m6623LrjypuErUduW++67LzfffHPKysrSpUuXTJw4McOGDVtnmVmzZuWmm25Kz549kyRdu3bNDTfcUES5vI8ZM2bkoYceSo8ePZIkAwf+//buPqbKuo/j+FvUkARJQlScnhzNrCyXOqshkDPMUpJ8qhAJJaVyFiwV1B6GEx9gKMEgGW6oKWbBGKFkKhvlbDSXcyTH8mFpPChCjczJSc6R+w/X2X3goKf7znOd2/vz2tw41+/7237XvvsevL5c1+8ysX79eoeYkydPkpGRgcViITAwkLVr1xIYGGjEcuU29u/fT3Fxsf3z1atXuXz5Mvv27bPXI7iWdzFOZ2cna9euJSQkhNjYWGw2G9nZ2dTU1GCz2Zg/fz6zZ8/uNs/VOHGvrvm0WCxkZmZiNpu5ceMGY8aMYcWKFfTr16/b3Li4OP7880/69Ln5pvtp06axYMECd5+CdNE1pwBTp05l0KBB9pgFCxYwbdq0bnOLioqorKzEZrMxbdo0Fi9erI3tDdY1n6mpqdTX19vHm5qaGDduHFlZWd3mupp3cY+erlNcrbs7UZ99/qvZ/6DQ0FBCQ0Mdjvn7+9sbEj3F3A3+et1pYWEhI0aMIDc3l7y8PFJSUv5WjHiGCxcukJOTwyeffEJgYCBHjx4lJSWFiooKh7gffviBpKQkfSn/D6itrSU9PZ3HH3/c6XhHRwepqamkp6czduxYSkpKWLduHdnZ2W5eqbhi+vTpTJ8+HQCr1cqSJUt47bXXHBoScPu8i3F+/vlnMjIyOHnyJCEhIQCUlZVRX1/Pnj17uHbtGgkJCYwePZpHH33UYa6rceI+zvJZVFSE1Wpl9+7ddHZ28uGHH7Jjxw4SExMd5ra3t9PQ0MDBgwftTQkxnrOcXrhwAT8/P4f/2ztz9OhRqqqq2LlzJ15eXrz99tuMHDmSyMhIdyxdnHCWz40bN9rHzWYzqamprFy5sttcV/Mu7tHTdUpqaqpLdXen6tMjHt/4f/fdd9/xyCOPMGLECABmz57NgQMH6Ozs/Fsx4hn69u3LmjVr7H8lf/jhh/n111/p6OhwiKutreWrr74iNjaWZcuWcfbsWSOWK7dx/fp1Tp8+za5du4iJiSElJYVLly45xJjNZvr378/YsWMBmDlzJseOHaOtrc2IJcvfsGPHDgICApg1a5bDcVfyLsYpKSkhKiqKZ5991n6surqaGTNm0KdPHwYMGEBkZCRffvllt7muxon7OMvnE088waJFi/Dy8qJ3796MGjWKixcvdptbV1fHvffeS3JyMq+++iqbN2/GYrG4c/nihLOc1tbW0rt3b958801iYmLYtm0bNput29zq6mqee+45fHx88Pb2JioqigMHDrhz+dKFs3z+paOjg7S0NJKTkxk8eHC3cVfzLu7R03VKVVWVS3V3p+pTTQkPoFei3l2Cg4OZNGkScPNWt+zsbMLDw+nbt689pr29HZPJxMKFC9m1axcvvvgiSUlJXLt2zahlSw9aWlqYMGECS5cuZffu3YwZM4bly5c7NASbm5sdfhH37duXgQMH0tLSYsSSxUVtbW0UFxeTnJzcbcyVvItxVqxYwQsvvOBwrGsdBgUFcfny5W5zXY0T93GWz6eeegqTyQTAxYsX+fTTT5kyZUq3udeuXWP8+PFs2LCB7du309zcTH5+vlvWLT1zllObzcbEiRP56KOPKCgooKamhs8++6zbXNWo53GWz7+Ul5cTGBjI5MmTnY67mndxj56uU1pbWw39HaqmhAfQK1HvTu3t7axatYqGhgbWrFnjMObj40Nubq79tvDIyEj8/Pwwm81GLFVuYdiwYWRnZ2MymejVqxexsbE0NjbS1NRkj+mpPr289BXrycrKyggPD2fYsGHdxlzJu3gWZ3XorAZdjRPPcOrUKZYsWcLcuXMJCwvrNh4eHk5aWhq+vr54e3sTHx9PdXW1+xcqtxUdHc3y5cu555578PPzIyYmxmmunDV/VaOea8+ePSxatKjHcVfzLu7V9TrF1bq7U/WpCvcAeiXq3efSpUskJCTQu3dv8vPz8fPzcxi/ePEie/fudTjW2dmp52E90JkzZ6isrHQ41jVXQ4YMobW11f7ZarXS1tbmcHeTeJ5Dhw4RFRXldMyVvItn6VqHLS0tTmvQ1Tgx3sGDB1m2bBlLly5l4cKFTmOOHDnC8ePH7Z9Vp56rsrKSM2fO2D/3lKvBgwc71Ojly5dVox7qp59+wmazMW7cuB5jXM27uI+z6xRX6+5O1aeaEh5Ar0S9u/z+++8kJiYyefJk0tPTne4U7uPjw9atW6mrqwNubhpjsVi00ZoH8vLyIisri8bGRgBKS0t58MEHHW5dGzNmDFeuXKG2thaAL774gscee6xbM0o8x5UrV2hoaOhxE0tX8i6eJTw8nIqKCqxWK3/88QeHDh3imWee+Y/jxFhVVVVkZWWRk5Nzyw2hm5ubycnJwWKxYLPZKC4udvrcuxjv3LlzFBQUYLPZsFgsfP755043x4uIiODAgQO0t7dz/fp19u/fT0REhAErlts5fvw4EyZMuOWbF1zNu7hHT9cprtbdnapPtak8QEBAAO+//z6pqakOrzs1m832V6L2FCOep7S0lObmZqqrqx1uT9uyZQvJyclkZ2czaNAg1q9fz4YNG+jo6KB///5kZGQ47DshniEkJITly5fz7rvvcuPGDYKCgli3bh0tLS0kJSXZ87lp0yYyMzNpb2/H399f9enhGhoaCAwMdPhrzb9/5/aUd/Fcs2fPprGxkfnz52O1WnnppZfsf70rKCgAIDEx8ZZx4jny8/Pp7OwkPT3dfmzs2LGsXLmS0tJSTp06xXvvvcesWbNoamoiLi4Om83G+PHjef311w1cufRk8eLFZGZmEhMTg9VqZcqUKcycORPAIadhYWGcPXuW+Ph4rFYr4eHh9jcmiWepr69n6NCh3Y7/+3furfIu7tfTdUpeXh6TJ092WnfuqM9ebW1t2rVLRERERERERNxOj2+IiIiIiIiIiCHUlBARERERERERQ6gpISIiIiIiIiKGUFNCRERERERERAyhpoSIiIiIiIiIGEKvBBUREZF/XFpaGvv37+9x/IMPPmDGjBluXJGIiIh4Ir0SVERERP5xV69exWKxAHDixAlWr15NZWWlfdzX15d+/foZtTwRERHxELpTQkRERP5xvr6++Pr6AjBgwAAAAgMDjVySiIiIeCDtKSEiIiJuN3HiRLZu3crUqVNZunQpcPOOivj4eMLCiIXr2QAAAwdJREFUwnj55Zcd7qwAKCsrIzo6moiICBYvXozZbLaPff/998TFxREWFkZUVBTbt2935+mIiIjIf0h3SoiIiIghvvnmGwoLC7FarbS2tpKcnExiYiKhoaH8+OOPbNiwAT8/P8LCwjhy5AgFBQWsWrWKkSNHcvjwYd566y1KSkoYOHAgKSkpzJs3j40bN3Lu3DlWr17NQw89xNNPP230aYqIiMgtqCkhIiIihoiOjsZkMgGwdetWxo8fzyuvvALA8OHDOX/+PHv27CEsLIydO3cSFxdHREQEAIsWLeLYsWOUl5czZ84crly5QkBAAMHBwQQHB5OXl0dwcLBh5yYiIiKuUVNCREREDDF06FD7z+fPn+fbb7+1Nx0AbDYb9913n338448/pqCgwD5+/fp1goKC8Pf3Z+7cuWRkZFBUVMSkSZN4/vnntYeFiIjI/wA1JURERMQQ3t7e9p+tVitTp04lISHBIcbL6+b2VzabjXfeeYcnn3zSYdzHxweAFStWMG/ePL7++muOHDnCG2+8wZo1a4iKirrDZyEiIiL/DW10KSIiIoYzmUzU19czfPhw+7+amhrKy8vt483NzQ7ju3fv5vjx47S2trJp0yaGDBlCXFwchYWFzJgxg8OHDxt8ViIiInI7akqIiIiI4ebMmcPp06fJy8vjl19+oaqqitzcXIKCggCIiYlh79697Nu3j4aGBgoLC6moqOCBBx7A39+f6upqNm/eTH19PXV1dZw4cYLRo0cbfFYiIiJyO3p8Q0RERAw3dOhQsrKyyMvLo7i4mPvvv58lS5YwZ84cACIjI/ntt9/Ytm0bra2tmEwmMjMzGTVqFABZWVls2bKF2NhYvL29iYyM7PYoiIiIiHieXm1tbZ1GL0JERERERERE/v/o8Q0RERERERERMYSaEiIiIiIiIiJiCDUlRERERERERMQQakqIiIiIiIiIiCHUlBARERERERERQ6gpISIiIiIiIiKGUFNCRERERERERAyhpoSIiIiIiIiIGEJNCRERERERERExxL8A7ysANeOpii8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1152x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "instance_depth_plot(X.sample(1),X.head(1),iForest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:39:39.813377Z",
     "start_time": "2020-06-14T17:39:39.806542Z"
    }
   },
   "outputs": [],
   "source": [
    "outlier  = evaluate_instance(X.head(1),iForest)\n",
    "normal  = evaluate_instance(X.sample(1),iForest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:39:47.292456Z",
     "start_time": "2020-06-14T17:39:47.287671Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4.15"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.mean(outlier)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:39:50.694882Z",
     "start_time": "2020-06-14T17:39:50.682671Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "10.35"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.mean(normal)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:43:12.950042Z",
     "start_time": "2020-06-14T17:43:12.944798Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2000, 2)"
      ]
     },
     "execution_count": 106,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:43:48.866947Z",
     "start_time": "2020-06-14T17:43:18.358917Z"
    }
   },
   "outputs": [],
   "source": [
    "iForest = isolation_forest(X,n_trees=20, max_depth=100, subspace=1_560)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-06-14T17:43:48.877794Z",
     "start_time": "2020-06-14T17:43:48.869247Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Anomaly score for outlier: 0.808515309012463\n",
      "Anomaly score for normal: 0.458312357739584\n"
     ]
    }
   ],
   "source": [
    "print('Anomaly score for outlier:',anomaly_score(X.head(1),iForest,1560))\n",
    "print('Anomaly score for normal:',anomaly_score(X.sample(1),iForest,1560))"
   ]
  },
  {
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}