Graph_Anomaly_Detection_Methods.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "a065614b-6432-49aa-a735-a268a67b587b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import networkx as nx\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.neighbors import LocalOutlierFactor\n",
    "from matplotlib_venn import venn2\n",
    "from node2vec import Node2Vec\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.model_selection import ParameterGrid\n",
    "from sklearn.ensemble import IsolationForest\n",
    "from sklearn import svm"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7ba3e240-d027-4bd2-9297-ae719049d83f",
   "metadata": {},
   "source": [
    "# Graph anomaly detection\n",
    "1. Generate labels for hub points using Local Outlier Factor according to internel attributes of graph such as degree centrality and betweenness centrality\n",
    "2. Compare the impact of degree centrality and betweenness centrality on the generated hub points\n",
    "3. Create a graph, where each edge has Utilization (boarded passengers concerning all departed flights' available seats), Regularity (percentage of operated flights concerning planned flights), ASM (available seat miles), RASM (Revenue Per Available Seat Mile) and Cancellation Probabilities as attributes\n",
    "4. Leverage Node2Vec technique to extract low-dimensional embeddingss of the graph using edge attributes\n",
    "5. Fine-tuning parameter p and q used in Node2Vec with a combination of outliers identified in step 1,2, LOF algorithm and F1-score.\n",
    "6. Implement One Class SVM and Isolation Random Forest on the node embeddings with best p, q to find the outliers\n",
    "7. Two-sample t test for examing metrics of the identified 'outliers'\n",
    "7. Link Prediction (Stratified Bagging with Logistic Regression, Random Forest Classifiers)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "110194c7-88aa-4512-b635-7c254ee609e9",
   "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>citypair</th>\n",
       "      <th>sum_departures_performed</th>\n",
       "      <th>sum_departures_scheduled</th>\n",
       "      <th>passengers</th>\n",
       "      <th>seats</th>\n",
       "      <th>avg_fuel_price</th>\n",
       "      <th>avg_stock_price</th>\n",
       "      <th>total_operating_expense</th>\n",
       "      <th>revenue</th>\n",
       "      <th>Aircraft_Unit_Cost ($ millions USD)</th>\n",
       "      <th>...</th>\n",
       "      <th>origin_density</th>\n",
       "      <th>dest_population</th>\n",
       "      <th>dest_density</th>\n",
       "      <th>DEP_DELAY_NEW</th>\n",
       "      <th>ARR_DELAY_NEW</th>\n",
       "      <th>CANCELLED</th>\n",
       "      <th>Utilization</th>\n",
       "      <th>Regularity</th>\n",
       "      <th>ASM</th>\n",
       "      <th>RASM</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ABQ-BWI</td>\n",
       "      <td>108.0</td>\n",
       "      <td>110.0</td>\n",
       "      <td>13736.0</td>\n",
       "      <td>15444.0</td>\n",
       "      <td>57.427544</td>\n",
       "      <td>52.842995</td>\n",
       "      <td>9585665.0</td>\n",
       "      <td>11058</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1155.5</td>\n",
       "      <td>2205092.0</td>\n",
       "      <td>2872.8</td>\n",
       "      <td>8.644068</td>\n",
       "      <td>9.440678</td>\n",
       "      <td>0.063492</td>\n",
       "      <td>0.889407</td>\n",
       "      <td>0.981818</td>\n",
       "      <td>9.246588</td>\n",
       "      <td>1.195901e+09</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ABQ-DAL</td>\n",
       "      <td>488.0</td>\n",
       "      <td>504.0</td>\n",
       "      <td>56344.0</td>\n",
       "      <td>74104.0</td>\n",
       "      <td>56.594778</td>\n",
       "      <td>53.008226</td>\n",
       "      <td>14229909.0</td>\n",
       "      <td>16207</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1155.5</td>\n",
       "      <td>5668165.0</td>\n",
       "      <td>1522.2</td>\n",
       "      <td>9.102740</td>\n",
       "      <td>8.075862</td>\n",
       "      <td>0.006803</td>\n",
       "      <td>0.760337</td>\n",
       "      <td>0.968254</td>\n",
       "      <td>127.715623</td>\n",
       "      <td>1.268991e+08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ABQ-DEN</td>\n",
       "      <td>330.0</td>\n",
       "      <td>337.0</td>\n",
       "      <td>37782.0</td>\n",
       "      <td>47510.0</td>\n",
       "      <td>56.594778</td>\n",
       "      <td>53.008226</td>\n",
       "      <td>14229909.0</td>\n",
       "      <td>16207</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1155.5</td>\n",
       "      <td>2650725.0</td>\n",
       "      <td>1805.7</td>\n",
       "      <td>18.624521</td>\n",
       "      <td>18.896154</td>\n",
       "      <td>0.003817</td>\n",
       "      <td>0.795243</td>\n",
       "      <td>0.979228</td>\n",
       "      <td>136.091530</td>\n",
       "      <td>1.190890e+08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ABQ-HOU</td>\n",
       "      <td>326.0</td>\n",
       "      <td>335.0</td>\n",
       "      <td>39590.0</td>\n",
       "      <td>50650.0</td>\n",
       "      <td>56.927884</td>\n",
       "      <td>52.942134</td>\n",
       "      <td>23815574.0</td>\n",
       "      <td>27265</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1155.5</td>\n",
       "      <td>5650910.0</td>\n",
       "      <td>1394.6</td>\n",
       "      <td>4.595506</td>\n",
       "      <td>4.044944</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.781639</td>\n",
       "      <td>0.973134</td>\n",
       "      <td>66.719030</td>\n",
       "      <td>4.086540e+08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ABQ-LAS</td>\n",
       "      <td>324.0</td>\n",
       "      <td>333.0</td>\n",
       "      <td>35966.0</td>\n",
       "      <td>46332.0</td>\n",
       "      <td>57.427544</td>\n",
       "      <td>52.842995</td>\n",
       "      <td>9585665.0</td>\n",
       "      <td>11058</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1155.5</td>\n",
       "      <td>2150373.0</td>\n",
       "      <td>1754.7</td>\n",
       "      <td>15.618321</td>\n",
       "      <td>13.946565</td>\n",
       "      <td>0.015038</td>\n",
       "      <td>0.776267</td>\n",
       "      <td>0.972973</td>\n",
       "      <td>95.250213</td>\n",
       "      <td>1.160942e+08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 36 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "  citypair  sum_departures_performed  sum_departures_scheduled  passengers  \\\n",
       "0  ABQ-BWI                     108.0                     110.0     13736.0   \n",
       "1  ABQ-DAL                     488.0                     504.0     56344.0   \n",
       "2  ABQ-DEN                     330.0                     337.0     37782.0   \n",
       "3  ABQ-HOU                     326.0                     335.0     39590.0   \n",
       "4  ABQ-LAS                     324.0                     333.0     35966.0   \n",
       "\n",
       "     seats  avg_fuel_price  avg_stock_price  total_operating_expense  revenue  \\\n",
       "0  15444.0       57.427544        52.842995                9585665.0    11058   \n",
       "1  74104.0       56.594778        53.008226               14229909.0    16207   \n",
       "2  47510.0       56.594778        53.008226               14229909.0    16207   \n",
       "3  50650.0       56.927884        52.942134               23815574.0    27265   \n",
       "4  46332.0       57.427544        52.842995                9585665.0    11058   \n",
       "\n",
       "   Aircraft_Unit_Cost ($ millions USD)  ...  origin_density  dest_population  \\\n",
       "0                                 74.0  ...          1155.5        2205092.0   \n",
       "1                                 74.0  ...          1155.5        5668165.0   \n",
       "2                                 74.0  ...          1155.5        2650725.0   \n",
       "3                                 74.0  ...          1155.5        5650910.0   \n",
       "4                                 74.0  ...          1155.5        2150373.0   \n",
       "\n",
       "   dest_density  DEP_DELAY_NEW ARR_DELAY_NEW CANCELLED Utilization  \\\n",
       "0        2872.8       8.644068      9.440678  0.063492    0.889407   \n",
       "1        1522.2       9.102740      8.075862  0.006803    0.760337   \n",
       "2        1805.7      18.624521     18.896154  0.003817    0.795243   \n",
       "3        1394.6       4.595506      4.044944  0.000000    0.781639   \n",
       "4        1754.7      15.618321     13.946565  0.015038    0.776267   \n",
       "\n",
       "   Regularity         ASM          RASM  \n",
       "0    0.981818    9.246588  1.195901e+09  \n",
       "1    0.968254  127.715623  1.268991e+08  \n",
       "2    0.979228  136.091530  1.190890e+08  \n",
       "3    0.973134   66.719030  4.086540e+08  \n",
       "4    0.972973   95.250213  1.160942e+08  \n",
       "\n",
       "[5 rows x 36 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data  = pd.read_csv('Database/FINAL_EDGE_TABLE.csv', index_col = 0)\n",
    "data = data.drop(columns = {'Unnamed: 0', 'Unnamed: 0.1', 'Unnamed: 0.2'}, axis = 1)\n",
    "# Utilization: boaded passengers concerning all departed flights’ available seats\n",
    "data = data.assign(Utilization = data['passengers'] / data['seats'])\n",
    "# Regularity: percentage of operated flights concerning planned flights\n",
    "data = data.assign(Regularity = data['sum_departures_performed'] / data['sum_departures_scheduled'])\n",
    "# Available seat miles (ASM): airline’s carrying capacity\n",
    "data = data.assign(ASM = data['seats'] / data['Route_Distance'])\n",
    "# Revenue Per Available Seat Mile (RASM): show financial performance\n",
    "data = data.assign(RASM = data['revenue'] * 1e06 / data['ASM'])\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f07bdb95-2e61-4a05-9b3a-92b5e951508b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['citypair', 'sum_departures_performed', 'sum_departures_scheduled',\n",
       "       'passengers', 'seats', 'avg_fuel_price', 'avg_stock_price',\n",
       "       'total_operating_expense', 'revenue',\n",
       "       'Aircraft_Unit_Cost ($ millions USD)',\n",
       "       'Aircraft_Inverse_Age 1/age (1/yrs)', 'Number_of_Cancellations',\n",
       "       'Proportion of Seats filled', 'Aircraft Age (yrs)',\n",
       "       'Origin_Airport_Code', 'Destination_Airport_Code',\n",
       "       'origin_airport_type', 'origin_lat', 'origin_long', 'origin_city',\n",
       "       'dest_airport_type', 'dest_lat', 'dest_long', 'dest_city',\n",
       "       'Route_Distance', 'origin_population', 'origin_density',\n",
       "       'dest_population', 'dest_density', 'DEP_DELAY_NEW', 'ARR_DELAY_NEW',\n",
       "       'CANCELLED', 'Utilization', 'Regularity', 'ASM', 'RASM'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "d8819363-730f-4132-9111-bf2804523732",
   "metadata": {},
   "outputs": [],
   "source": [
    "data.to_csv('airline_tab.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "81026ac7-838f-4ec6-a494-6ffda4a182ce",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 6 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data[['Utilization', 'Regularity', 'ASM', 'RASM', 'CANCELLED']].hist(layout = (2,3));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "c69d6b3a-5aa7-4f21-9273-c595c4ff2068",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create a graph\n",
    "G = nx.from_pandas_edgelist(data, source = 'Origin_Airport_Code', target = 'Destination_Airport_Code', create_using = nx.DiGraph, edge_attr = ['Utilization', 'Regularity', 'CANCELLED', 'ASM', 'RASM'])\n",
    "# Calculate degree centrality for each airport\n",
    "degree_centrality = nx.degree_centrality(G)\n",
    "# Calculate betweenness centrality for each airport\n",
    "betweenness_centrality = nx.betweenness_centrality(G)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "f60310a1-695f-4b8f-a315-4abd1b5a7767",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get the nodes and edges of the graph\n",
    "nodes = [node for node in G.nodes()]\n",
    "edges = [edge for edge in G.edges()]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf86e8e1-37d1-4226-af94-513fad1900ec",
   "metadata": {
    "tags": []
   },
   "source": [
    "## Create lables for outliers (hub points)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "71915308-2d12-4c87-8d9c-4956076032f6",
   "metadata": {},
   "source": [
    "### Use degree centrality as the attribute"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "581bd53b-b29a-4095-b3ac-929e0b10aa13",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--LOF with Degree Centrality--\n",
      "Outliers are ['BWI', 'DAL', 'DEN', 'HOU', 'LAS', 'LAX', 'MDW', 'OAK', 'PHX', 'MCO', 'TPA', 'BNA', 'CMH', 'DCA', 'IND', 'MKE', 'MSY', 'SAN', 'SAT', 'STL', 'SJC', 'SMF']\n"
     ]
    }
   ],
   "source": [
    "X = np.array(list(degree_centrality.values())).reshape(-1, 1)\n",
    "lof_deg = LocalOutlierFactor(n_neighbors = 20).fit(X)\n",
    "# Nodes with LOF greater than 1.5 are identified as outliers (i.e negative outlier factor < -1.5)\n",
    "outliers_deg = [nodes[i] for i in np.where(lof_deg.negative_outlier_factor_ < lof_deg.offset_)[0]]\n",
    "# Create labels (1: inliers, -1: outliers)\n",
    "y_deg = np.array([-1 if i in outliers_deg else 1 for i in nodes])\n",
    "print('--LOF with Degree Centrality--')\n",
    "print('Outliers are {}'.format(outliers_deg))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d1a12162-4343-408c-9c7a-a805cbdc78be",
   "metadata": {},
   "source": [
    "### Use betweenness centrality as the attribute"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "8cf72d4b-d3ea-4125-8398-107cdacee111",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--LOF with Betweenness Centrality--\n",
      "Outliers are ['ABQ', 'BWI', 'DAL', 'DEN', 'HOU', 'LAS', 'LAX', 'MDW', 'OAK', 'PHX', 'ALB', 'FLL', 'MCO', 'TPA', 'ATL', 'AUS', 'BNA', 'BOS', 'CLE', 'CMH', 'DCA', 'IND', 'LGA', 'MCI', 'MKE', 'MSP', 'MSY', 'PBI', 'PHL', 'PIT', 'RSW', 'SAN', 'SAT', 'STL', 'ELP', 'EWR', 'SFO', 'SJC', 'SMF', 'BDL', 'BHM', 'RDU', 'BOI', 'GEG', 'BUF', 'BUR', 'PDX', 'SLC', 'GRR', 'ISP', 'MEM', 'MHT', 'OKC', 'PVD', 'ROC', 'SDF', 'SJU', 'OMA', 'SEA', 'SNA', 'ONT', 'RNO', 'TUS', 'LGB']\n"
     ]
    }
   ],
   "source": [
    "X = np.array(list(betweenness_centrality.values())).reshape(-1, 1)\n",
    "lof_betw = LocalOutlierFactor(n_neighbors = 20).fit(X)\n",
    "# Nodes with LOF greater than 1.5 are identified as outliers (i.e negative outlier factor < -1.5)\n",
    "outliers_betw = [nodes[i] for i in np.where(lof_betw.negative_outlier_factor_ < lof_betw.offset_)[0]]\n",
    "# Create labels (1: inliers, -1: outliers)\n",
    "y_betw = np.array([-1 if i in outliers_betw else 1 for i in nodes])\n",
    "print('--LOF with Betweenness Centrality--')\n",
    "print('Outliers are {}'.format(outliers_betw))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5b3fb1db-8c6b-4593-9a1d-71f3e8936e84",
   "metadata": {},
   "source": [
    "### Compare the outliers identified with Degree Centrality and Betweenness Centrality"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "1e92d687-8bd6-41ad-bb73-b6ef40e10249",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib_venn import venn2\n",
    "\n",
    "venn = venn2([set(outliers_deg), set(outliers_betw)], set_labels=('Degree Centrality', 'Betweenness Centrality'), set_colors=('green', 'yellow'), alpha = 0.7)\n",
    "plt.legend(('Degree Centrality', 'Betweenness Centrality'), loc='center left', bbox_to_anchor=(1, 0.5), frameon = False, prop = {'size': 10}, markerscale = 1.5, title = 'Internal Attribute of the Network', title_fontsize = 12, handletextpad = 0.2)\n",
    "plt.title('Venn Diagram of Number of Outliers')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "16548898-86a3-4011-ac1d-eeb9bbb17d58",
   "metadata": {},
   "source": [
    "#### All outliers recognized using degree centrality belong to those identified using betweenness centrality"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "614579ac-7fe0-48d0-9e58-8ff29057445a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of Common is: 22\n",
      "Number of Unique Outliers obtained using Betweenness Centrality is: 42\n"
     ]
    }
   ],
   "source": [
    "# Find the common outliers\n",
    "common_outliers = set(outliers_deg).intersection(outliers_betw)\n",
    "# Find the unique outliers identified using betweenness centrality\n",
    "unique_outliers_betw = [i for i in outliers_betw if i not in common_outliers]\n",
    "print(\"Number of Common is: {}\".format(len(outliers_deg)))\n",
    "print(\"Number of Unique Outliers obtained using Betweenness Centrality is: {}\".format(len(unique_outliers_betw)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9bd77e7e-aad7-434d-9b89-cca896b2d109",
   "metadata": {},
   "source": [
    "### Use LOF to identify anomaly edges based on Utilization, Regularity, CANCELLED, ASM, RASM metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "b565b592-13bf-4a09-96d9-8d5c2263ca39",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of Edges identified as outliers: 30\n"
     ]
    }
   ],
   "source": [
    "from sklearn.preprocessing import MinMaxScaler, StandardScaler\n",
    "# Normalize the features so that they are on the same scale\n",
    "X = data.loc[:, ['Utilization', 'Regularity', 'CANCELLED', 'ASM', 'RASM']]\n",
    "scaler = StandardScaler()\n",
    "X = scaler.fit_transform(X)\n",
    "\n",
    "# Compute the LOF scores based on the metrics\n",
    "lof = LocalOutlierFactor(n_neighbors = 10)\n",
    "lof_scores = lof.fit_predict(X)\n",
    "edge_outliers = [edges[i] for i in np.where(lof.negative_outlier_factor_ < lof.offset_)[0]]\n",
    "print(f\"The number of Edges identified as outliers: {len(edge_outliers)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "5e8d39e9-8e08-419e-b09b-e390b9757083",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('BWI', 'SMF'),\n",
       " ('DEN', 'GRR'),\n",
       " ('DEN', 'MCO'),\n",
       " ('HOU', 'SNA'),\n",
       " ('LAS', 'MKE'),\n",
       " ('LAS', 'ONT'),\n",
       " ('LAX', 'BNA'),\n",
       " ('LAX', 'SFO'),\n",
       " ('MDW', 'BWI'),\n",
       " ('PHX', 'CMH'),\n",
       " ('PHX', 'DEN'),\n",
       " ('PHX', 'ELP'),\n",
       " ('FLL', 'IND'),\n",
       " ('MCO', 'ATL'),\n",
       " ('MCO', 'AUS'),\n",
       " ('MCO', 'PHL'),\n",
       " ('BNA', 'CLE'),\n",
       " ('DCA', 'STL'),\n",
       " ('DTW', 'STL'),\n",
       " ('PHL', 'HOU'),\n",
       " ('PHL', 'MCO'),\n",
       " ('RSW', 'BNA'),\n",
       " ('STL', 'RSW'),\n",
       " ('ELP', 'HOU'),\n",
       " ('HRL', 'AUS'),\n",
       " ('HRL', 'HOU'),\n",
       " ('SJC', 'HOU'),\n",
       " ('SJC', 'PHX'),\n",
       " ('BDL', 'TPA'),\n",
       " ('PDX', 'STL')]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "edge_outliers"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c0a3408e-621c-4c6f-aed9-c5aee41373a5",
   "metadata": {},
   "source": [
    "## Fine-tuning p q in Node2Vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "2ca7df5d-1a7e-47ba-a7c5-5a52cecab535",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.1, 'q': 0.25}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "974b2491d8814366bcf6c16e258e6429",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 87.17it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.1, 'q': 0.5}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e65967a4704d42a1a3f6745ca71da5c0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 83.44it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.1, 'q': 0.75}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "101448ca97cf457389c83861dd96414d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 83.03it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.1, 'q': 0.9}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b688f46e87904479bb714ca86fb125e2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 81.06it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.1, 'q': 1}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4604d32c158c400c9363dc81808ef760",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 78.50it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.25, 'q': 0.5}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "69e60f2fe91e4416b3bb9b960f2c0009",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 80.74it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.25, 'q': 0.75}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1f37162724cb421fa4992e9cdad3b7e4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 84.52it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.25, 'q': 0.9}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bf2ed9bee6aa48e1a377e7ab602f0928",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 75.81it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.25, 'q': 1}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b287d32cdf9846748c8721ea30cce3c5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 82.73it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.5, 'q': 0.75}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2666dbc6ade443e9bb957111b3152219",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 82.05it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.5, 'q': 0.9}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "310f746a3fc44817b76498f5a4cafdfb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 81.53it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.5, 'q': 1}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1d208751281045c7950be53b3126bf4a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 82.69it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.75, 'q': 0.9}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f2375059ab2647fb8dbb6c2544f3449d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 82.50it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.75, 'q': 1}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4976667d3b14494aa6fd0f01657d23de",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 80.80it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'p': 0.9, 'q': 1}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "919ae54411014d75bb8caee03196f62d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Computing transition probabilities:   0%|          | 0/85 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Generating walks (CPU: 1): 100%|██████████| 100/100 [00:01<00:00, 78.99it/s]\n"
     ]
    }
   ],
   "source": [
    "# Define the hyperparameters to search over\n",
    "hyperparams = {\n",
    "    'p': [0.1, 0.25, 0.5, 0.75, 0.9, 1],\n",
    "    'q': [0.1, 0.25, 0.5, 0.75, 0.9, 1]\n",
    "}\n",
    "best_p_q = dict()\n",
    "\n",
    "# Generate all possible combinations of hyperparameters\n",
    "param_grid = ParameterGrid(hyperparams)\n",
    "# Initialize the evaluation metric\n",
    "best_f1_score = -1\n",
    "# Iterate over each combination of hyperparameters\n",
    "for params in param_grid:\n",
    "    if params['p'] >= params['q']:\n",
    "        continue\n",
    "    print(params)\n",
    "    # Initialize a Node2Vec object with the weighted graph and the weight keys\n",
    "    node2vec = Node2Vec(G, dimensions = 32, walk_length = 30, num_walks = 100,  p = params['p'], q = params['q'], weight_key = ('Utilization', 'Regularity', 'CANCELLED', 'ASM', 'RASM'))\n",
    "    # Train the node2vec model\n",
    "    model = node2vec.fit(window = 10, min_count = 1)\n",
    "    # Get the node embeddings\n",
    "    embeddings = np.zeros((G.number_of_nodes(), model.vector_size))\n",
    "    for i, node in enumerate(G.nodes()):\n",
    "        embeddings[i] = model.wv[node]\n",
    "    # Apply LOF to predict outlies based on the embeddings\n",
    "        lof_n2v = LocalOutlierFactor(n_neighbors = 20).fit(embeddings)\n",
    "        outliers_n2v = [nodes[i] for i in np.where(lof_n2v.negative_outlier_factor_ < lof_n2v.offset_)[0]]\n",
    "        y_pred_lof = np.array([-1 if i in outliers_n2v else 1 for i in nodes])\n",
    "        f1 = f1_score(y_deg, y_pred_lof)\n",
    "        if f1 >= best_f1_score:\n",
    "            best_f1_score = f1\n",
    "            best_p_q['p'], best_p_q['q']  = params['p'], params['q']\n",
    "            node_embeddings = embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "8d00db3c-2177-4f74-83de-9677b59d8245",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'p': 0.9, 'q': 1}"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_p_q"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "63ca095e-0a05-4b33-83d8-3aaa1e327c99",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.898550724637681"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_f1_score"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2ff0bccc-af5e-4a9f-93ba-faa575828c2c",
   "metadata": {},
   "source": [
    "## IsolationRandomForest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "eebef387-54e4-47cd-a19b-5dcaf4ff2bd3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# create an IsolationForest object\n",
    "isolation_forest = IsolationForest(n_estimators = 100, contamination = 0.15, random_state = 42)\n",
    "\n",
    "# fit the IsolationForest model to the embeddings\n",
    "isolation_forest.fit(node_embeddings)\n",
    "# get the average anomaly\n",
    "outlier_scores_isof = isolation_forest.decision_function(node_embeddings)\n",
    "\n",
    "plt.figure(figsize=(8, 6))\n",
    "plt.hist(outlier_scores_isof, bins = 50)\n",
    "plt.xlabel(\"outlier score\")\n",
    "plt.ylabel(\"Frequency\")\n",
    "plt.title(\"Isolation Random Forest\");"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "51462cb9-8537-4a91-b1ce-842515da3fd7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Anomaly nodes: ['GSP', 'PBI', 'RIC', 'RSW', 'HRL', 'SMF', 'BOI', 'GEG', 'ISP', 'PWM', 'CRP', 'DSM', 'LGB']\n"
     ]
    }
   ],
   "source": [
    "# Identify the anomaly nodes based on positiveness of the scores (inliers: positive score; outliers: negative score)\n",
    "outliers_isof = [nodes[i] for i in np.where(outlier_scores_isof < 0)[0]]\n",
    "print(\"Anomaly nodes:\", outliers_isof)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "57ed7137-8033-4eaa-92fd-c3a71ff23f16",
   "metadata": {},
   "source": [
    "### One class SVM \n",
    "Since we don't have the labels for the airport such that we can assume all nodes belongs to only one class. OC-SVM uses the origin as the other class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5a136d06-c45e-40de-8d95-d616ef00b4fb",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Train the One-Class SVM\n",
    "svm = svm.OneClassSVM(nu = 0.5, kernel = \"rbf\", gamma = 0.1)\n",
    "svm.fit(node_embeddings)\n",
    "\n",
    "# Predict the anomaly scores for all embeddings\n",
    "outlier_scores_svm = svm.decision_function(node_embeddings)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "46be5fa4-d9db-4637-82a1-6c0c0b52f817",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(outlier_scores_svm, bins = 50)\n",
    "plt.xlabel(\"outlier score\")\n",
    "plt.ylabel(\"Frequency\")\n",
    "plt.title(\"One Class SVM\");"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "349834cf-59b5-4eb8-9aa9-58c1864624ad",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Anomaly nodes: ['BWI', 'HOU', 'LAS', 'OAK', 'FLL', 'TPA', 'AMA', 'ATL', 'AUS', 'BNA', 'DCA', 'GSP', 'IAD', 'PBI', 'RIC', 'RSW', 'STL', 'HRL', 'LBB', 'SJC', 'SMF', 'BDL', 'PNS', 'BOI', 'GEG', 'BUR', 'CVG', 'GRR', 'ISP', 'MHT', 'PVD', 'PWM', 'ROC', 'SJU', 'CRP', 'LIT', 'MAF', 'SNA', 'ONT', 'TUS', 'DSM', 'ICT', 'LGB']\n"
     ]
    }
   ],
   "source": [
    "# Identify nodes as anomalies if its anomaly scoce is below zero\n",
    "outliers_svm = [nodes[i] for i in np.where(outlier_scores_svm < 0)[0]]\n",
    "print(\"Anomaly nodes:\", outliers_svm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "01b17c4c-752f-4b31-97db-28bb002aef7a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'BOI',\n",
       " 'CRP',\n",
       " 'DSM',\n",
       " 'GEG',\n",
       " 'GSP',\n",
       " 'HRL',\n",
       " 'ISP',\n",
       " 'LGB',\n",
       " 'PBI',\n",
       " 'PWM',\n",
       " 'RIC',\n",
       " 'RSW',\n",
       " 'SMF'}"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "potential_outliers = set(outliers_isof).intersection(outliers_svm)\n",
    "potential_outliers"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e557693e-66ba-45b4-aa02-93127cd94c4e",
   "metadata": {},
   "source": [
    "### Two-sample t-test with unequal variances for removing potential hub points identified using graph anomaly detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "21e1e190-bda1-4760-81b4-ac51669f6131",
   "metadata": {},
   "outputs": [],
   "source": [
    "from scipy import stats\n",
    "def ttest_removing_potential_hubs(data, metric, node):\n",
    "    t_stat, p_val = stats.ttest_ind(data[data['citypair'].str.contains(node)][metric], data[~data['citypair'].str.contains(node)][metric], equal_var = False)\n",
    "    return p_val"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "67332727-6768-49c0-998c-40ce102b4002",
   "metadata": {},
   "outputs": [],
   "source": [
    "uncertain_nodes = [node for node in potential_outliers if len(data[data['citypair'].str.contains(node)]) <= 2]\n",
    "new_candidates = [node for node in potential_outliers if node not in uncertain_nodes]\n",
    "metrics = ['CANCELLED', 'Utilization', 'Regularity', 'ASM', 'RASM']\n",
    "res = {}\n",
    "for node in new_candidates:\n",
    "    for metric in metrics:\n",
    "        p_val = ttest_removing_potential_hubs(data, metric, node)\n",
    "        if p_val > 0.05:\n",
    "            if node not in res:\n",
    "                res[node] = [(metric, p_val)]\n",
    "            else:\n",
    "                res[node].append((metric, p_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "098ad761-2f1c-4014-97c7-bd8306443fb3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'ISP': [('CANCELLED', 0.11493770284952684), ('ASM', 0.8708125408861225)],\n",
       " 'LGB': [('CANCELLED', 0.6126524563822475),\n",
       "  ('Utilization', 0.08271168439184777),\n",
       "  ('Regularity', 0.11109837099748708),\n",
       "  ('ASM', 0.4827951444934456)],\n",
       " 'GEG': [('CANCELLED', 0.6152638910356661)],\n",
       " 'BOI': [('Utilization', 0.4047586017538669),\n",
       "  ('Regularity', 0.7794564866150244),\n",
       "  ('RASM', 0.6864971537246725)],\n",
       " 'SMF': [('Utilization', 0.19683332562939587),\n",
       "  ('Regularity', 0.2728959081685385),\n",
       "  ('ASM', 0.0737441818263532),\n",
       "  ('RASM', 0.1819152132500021)],\n",
       " 'PBI': [('Utilization', 0.2300058421799936)],\n",
       " 'DSM': [('Utilization', 0.9260569774811551),\n",
       "  ('Regularity', 0.903384018200976),\n",
       "  ('ASM', 0.1313361123195673),\n",
       "  ('RASM', 0.47428663919142017)],\n",
       " 'HRL': [('Regularity', 0.3778551269229868), ('ASM', 0.5086221437609962)],\n",
       " 'RSW': [('Regularity', 0.15939838840123347), ('RASM', 0.6643025913571261)]}"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "d785c6b2-a971-4367-bca6-d002e47f34e2",
   "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>citypair</th>\n",
       "      <th>sum_departures_performed</th>\n",
       "      <th>sum_departures_scheduled</th>\n",
       "      <th>passengers</th>\n",
       "      <th>seats</th>\n",
       "      <th>avg_fuel_price</th>\n",
       "      <th>avg_stock_price</th>\n",
       "      <th>total_operating_expense</th>\n",
       "      <th>revenue</th>\n",
       "      <th>Aircraft_Unit_Cost ($ millions USD)</th>\n",
       "      <th>...</th>\n",
       "      <th>origin_density</th>\n",
       "      <th>dest_population</th>\n",
       "      <th>dest_density</th>\n",
       "      <th>DEP_DELAY_NEW</th>\n",
       "      <th>ARR_DELAY_NEW</th>\n",
       "      <th>CANCELLED</th>\n",
       "      <th>Utilization</th>\n",
       "      <th>Regularity</th>\n",
       "      <th>ASM</th>\n",
       "      <th>RASM</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>ATL-RIC</td>\n",
       "      <td>310.0</td>\n",
       "      <td>315.0</td>\n",
       "      <td>32691.0</td>\n",
       "      <td>44330.0</td>\n",
       "      <td>57.427544</td>\n",
       "      <td>52.842995</td>\n",
       "      <td>9585665.0</td>\n",
       "      <td>11058</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1419.9</td>\n",
       "      <td>1008069.0</td>\n",
       "      <td>1477.0</td>\n",
       "      <td>10.943005</td>\n",
       "      <td>10.227979</td>\n",
       "      <td>0.007712</td>\n",
       "      <td>0.737446</td>\n",
       "      <td>0.984127</td>\n",
       "      <td>92.189936</td>\n",
       "      <td>1.199480e+08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>996</th>\n",
       "      <td>RIC-ATL</td>\n",
       "      <td>308.0</td>\n",
       "      <td>315.0</td>\n",
       "      <td>31788.0</td>\n",
       "      <td>44044.0</td>\n",
       "      <td>57.427544</td>\n",
       "      <td>52.842995</td>\n",
       "      <td>9585665.0</td>\n",
       "      <td>11058</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1477.0</td>\n",
       "      <td>5046555.0</td>\n",
       "      <td>1419.9</td>\n",
       "      <td>6.626943</td>\n",
       "      <td>6.821244</td>\n",
       "      <td>0.005155</td>\n",
       "      <td>0.721733</td>\n",
       "      <td>0.977778</td>\n",
       "      <td>91.595162</td>\n",
       "      <td>1.207269e+08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 36 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    citypair  sum_departures_performed  sum_departures_scheduled  passengers  \\\n",
       "47   ATL-RIC                     310.0                     315.0     32691.0   \n",
       "996  RIC-ATL                     308.0                     315.0     31788.0   \n",
       "\n",
       "       seats  avg_fuel_price  avg_stock_price  total_operating_expense  \\\n",
       "47   44330.0       57.427544        52.842995                9585665.0   \n",
       "996  44044.0       57.427544        52.842995                9585665.0   \n",
       "\n",
       "     revenue  Aircraft_Unit_Cost ($ millions USD)  ...  origin_density  \\\n",
       "47     11058                                 74.0  ...          1419.9   \n",
       "996    11058                                 74.0  ...          1477.0   \n",
       "\n",
       "     dest_population  dest_density  DEP_DELAY_NEW ARR_DELAY_NEW CANCELLED  \\\n",
       "47         1008069.0        1477.0      10.943005     10.227979  0.007712   \n",
       "996        5046555.0        1419.9       6.626943      6.821244  0.005155   \n",
       "\n",
       "    Utilization  Regularity        ASM          RASM  \n",
       "47     0.737446    0.984127  92.189936  1.199480e+08  \n",
       "996    0.721733    0.977778  91.595162  1.207269e+08  \n",
       "\n",
       "[2 rows x 36 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "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>citypair</th>\n",
       "      <th>sum_departures_performed</th>\n",
       "      <th>sum_departures_scheduled</th>\n",
       "      <th>passengers</th>\n",
       "      <th>seats</th>\n",
       "      <th>avg_fuel_price</th>\n",
       "      <th>avg_stock_price</th>\n",
       "      <th>total_operating_expense</th>\n",
       "      <th>revenue</th>\n",
       "      <th>Aircraft_Unit_Cost ($ millions USD)</th>\n",
       "      <th>...</th>\n",
       "      <th>origin_density</th>\n",
       "      <th>dest_population</th>\n",
       "      <th>dest_density</th>\n",
       "      <th>DEP_DELAY_NEW</th>\n",
       "      <th>ARR_DELAY_NEW</th>\n",
       "      <th>CANCELLED</th>\n",
       "      <th>Utilization</th>\n",
       "      <th>Regularity</th>\n",
       "      <th>ASM</th>\n",
       "      <th>RASM</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>203</th>\n",
       "      <td>BWI-PWM</td>\n",
       "      <td>273.0</td>\n",
       "      <td>284.0</td>\n",
       "      <td>29618.0</td>\n",
       "      <td>39647.0</td>\n",
       "      <td>56.594778</td>\n",
       "      <td>53.008226</td>\n",
       "      <td>14229909.0</td>\n",
       "      <td>16207</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2872.8</td>\n",
       "      <td>2036875.0</td>\n",
       "      <td>1881.6</td>\n",
       "      <td>10.378151</td>\n",
       "      <td>10.680672</td>\n",
       "      <td>0.03252</td>\n",
       "      <td>0.747043</td>\n",
       "      <td>0.961268</td>\n",
       "      <td>87.757762</td>\n",
       "      <td>1.846788e+08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>982</th>\n",
       "      <td>PWM-BWI</td>\n",
       "      <td>270.0</td>\n",
       "      <td>283.0</td>\n",
       "      <td>31110.0</td>\n",
       "      <td>39218.0</td>\n",
       "      <td>56.594778</td>\n",
       "      <td>53.008226</td>\n",
       "      <td>14229909.0</td>\n",
       "      <td>16207</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1881.6</td>\n",
       "      <td>2205092.0</td>\n",
       "      <td>2872.8</td>\n",
       "      <td>14.655738</td>\n",
       "      <td>14.590164</td>\n",
       "      <td>0.00813</td>\n",
       "      <td>0.793258</td>\n",
       "      <td>0.954064</td>\n",
       "      <td>86.808180</td>\n",
       "      <td>1.866990e+08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 36 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    citypair  sum_departures_performed  sum_departures_scheduled  passengers  \\\n",
       "203  BWI-PWM                     273.0                     284.0     29618.0   \n",
       "982  PWM-BWI                     270.0                     283.0     31110.0   \n",
       "\n",
       "       seats  avg_fuel_price  avg_stock_price  total_operating_expense  \\\n",
       "203  39647.0       56.594778        53.008226               14229909.0   \n",
       "982  39218.0       56.594778        53.008226               14229909.0   \n",
       "\n",
       "     revenue  Aircraft_Unit_Cost ($ millions USD)  ...  origin_density  \\\n",
       "203    16207                                 74.0  ...          2872.8   \n",
       "982    16207                                 74.0  ...          1881.6   \n",
       "\n",
       "     dest_population  dest_density  DEP_DELAY_NEW ARR_DELAY_NEW CANCELLED  \\\n",
       "203        2036875.0        1881.6      10.378151     10.680672   0.03252   \n",
       "982        2205092.0        2872.8      14.655738     14.590164   0.00813   \n",
       "\n",
       "    Utilization  Regularity        ASM          RASM  \n",
       "203    0.747043    0.961268  87.757762  1.846788e+08  \n",
       "982    0.793258    0.954064  86.808180  1.866990e+08  \n",
       "\n",
       "[2 rows x 36 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "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>citypair</th>\n",
       "      <th>sum_departures_performed</th>\n",
       "      <th>sum_departures_scheduled</th>\n",
       "      <th>passengers</th>\n",
       "      <th>seats</th>\n",
       "      <th>avg_fuel_price</th>\n",
       "      <th>avg_stock_price</th>\n",
       "      <th>total_operating_expense</th>\n",
       "      <th>revenue</th>\n",
       "      <th>Aircraft_Unit_Cost ($ millions USD)</th>\n",
       "      <th>...</th>\n",
       "      <th>origin_density</th>\n",
       "      <th>dest_population</th>\n",
       "      <th>dest_density</th>\n",
       "      <th>DEP_DELAY_NEW</th>\n",
       "      <th>ARR_DELAY_NEW</th>\n",
       "      <th>CANCELLED</th>\n",
       "      <th>Utilization</th>\n",
       "      <th>Regularity</th>\n",
       "      <th>ASM</th>\n",
       "      <th>RASM</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>249</th>\n",
       "      <td>CRP-HOU</td>\n",
       "      <td>418.0</td>\n",
       "      <td>435.0</td>\n",
       "      <td>30540.0</td>\n",
       "      <td>59806.0</td>\n",
       "      <td>58.260309</td>\n",
       "      <td>52.677764</td>\n",
       "      <td>14527086.0</td>\n",
       "      <td>16967</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>777.0</td>\n",
       "      <td>5650910.0</td>\n",
       "      <td>1394.6</td>\n",
       "      <td>5.844828</td>\n",
       "      <td>7.086207</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.510651</td>\n",
       "      <td>0.96092</td>\n",
       "      <td>320.24772</td>\n",
       "      <td>5.298086e+07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>463</th>\n",
       "      <td>HOU-CRP</td>\n",
       "      <td>418.0</td>\n",
       "      <td>435.0</td>\n",
       "      <td>30170.0</td>\n",
       "      <td>59806.0</td>\n",
       "      <td>58.260309</td>\n",
       "      <td>52.677764</td>\n",
       "      <td>14527086.0</td>\n",
       "      <td>16967</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1394.6</td>\n",
       "      <td>333764.0</td>\n",
       "      <td>777.0</td>\n",
       "      <td>12.732759</td>\n",
       "      <td>11.612069</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.504464</td>\n",
       "      <td>0.96092</td>\n",
       "      <td>320.24772</td>\n",
       "      <td>5.298086e+07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 36 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    citypair  sum_departures_performed  sum_departures_scheduled  passengers  \\\n",
       "249  CRP-HOU                     418.0                     435.0     30540.0   \n",
       "463  HOU-CRP                     418.0                     435.0     30170.0   \n",
       "\n",
       "       seats  avg_fuel_price  avg_stock_price  total_operating_expense  \\\n",
       "249  59806.0       58.260309        52.677764               14527086.0   \n",
       "463  59806.0       58.260309        52.677764               14527086.0   \n",
       "\n",
       "     revenue  Aircraft_Unit_Cost ($ millions USD)  ...  origin_density  \\\n",
       "249    16967                                 74.0  ...           777.0   \n",
       "463    16967                                 74.0  ...          1394.6   \n",
       "\n",
       "     dest_population  dest_density  DEP_DELAY_NEW ARR_DELAY_NEW CANCELLED  \\\n",
       "249        5650910.0        1394.6       5.844828      7.086207       0.0   \n",
       "463         333764.0         777.0      12.732759     11.612069       0.0   \n",
       "\n",
       "    Utilization  Regularity        ASM          RASM  \n",
       "249    0.510651     0.96092  320.24772  5.298086e+07  \n",
       "463    0.504464     0.96092  320.24772  5.298086e+07  \n",
       "\n",
       "[2 rows x 36 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "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>citypair</th>\n",
       "      <th>sum_departures_performed</th>\n",
       "      <th>sum_departures_scheduled</th>\n",
       "      <th>passengers</th>\n",
       "      <th>seats</th>\n",
       "      <th>avg_fuel_price</th>\n",
       "      <th>avg_stock_price</th>\n",
       "      <th>total_operating_expense</th>\n",
       "      <th>revenue</th>\n",
       "      <th>Aircraft_Unit_Cost ($ millions USD)</th>\n",
       "      <th>...</th>\n",
       "      <th>origin_density</th>\n",
       "      <th>dest_population</th>\n",
       "      <th>dest_density</th>\n",
       "      <th>DEP_DELAY_NEW</th>\n",
       "      <th>ARR_DELAY_NEW</th>\n",
       "      <th>CANCELLED</th>\n",
       "      <th>Utilization</th>\n",
       "      <th>Regularity</th>\n",
       "      <th>ASM</th>\n",
       "      <th>RASM</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>ATL-GSP</td>\n",
       "      <td>239.0</td>\n",
       "      <td>241.0</td>\n",
       "      <td>21988.0</td>\n",
       "      <td>34177.0</td>\n",
       "      <td>54.929246</td>\n",
       "      <td>53.338689</td>\n",
       "      <td>4644244.0</td>\n",
       "      <td>5149</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>1419.9</td>\n",
       "      <td>35151.0</td>\n",
       "      <td>528.5</td>\n",
       "      <td>13.302752</td>\n",
       "      <td>12.614679</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.643357</td>\n",
       "      <td>0.991701</td>\n",
       "      <td>222.893544</td>\n",
       "      <td>2.310071e+07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>452</th>\n",
       "      <td>GSP-ATL</td>\n",
       "      <td>239.0</td>\n",
       "      <td>241.0</td>\n",
       "      <td>21729.0</td>\n",
       "      <td>34177.0</td>\n",
       "      <td>54.929246</td>\n",
       "      <td>53.338689</td>\n",
       "      <td>4644244.0</td>\n",
       "      <td>5149</td>\n",
       "      <td>74.0</td>\n",
       "      <td>...</td>\n",
       "      <td>528.5</td>\n",
       "      <td>5046555.0</td>\n",
       "      <td>1419.9</td>\n",
       "      <td>8.532308</td>\n",
       "      <td>8.864198</td>\n",
       "      <td>0.006116</td>\n",
       "      <td>0.635778</td>\n",
       "      <td>0.991701</td>\n",
       "      <td>222.893544</td>\n",
       "      <td>2.310071e+07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 36 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    citypair  sum_departures_performed  sum_departures_scheduled  passengers  \\\n",
       "28   ATL-GSP                     239.0                     241.0     21988.0   \n",
       "452  GSP-ATL                     239.0                     241.0     21729.0   \n",
       "\n",
       "       seats  avg_fuel_price  avg_stock_price  total_operating_expense  \\\n",
       "28   34177.0       54.929246        53.338689                4644244.0   \n",
       "452  34177.0       54.929246        53.338689                4644244.0   \n",
       "\n",
       "     revenue  Aircraft_Unit_Cost ($ millions USD)  ...  origin_density  \\\n",
       "28      5149                                 74.0  ...          1419.9   \n",
       "452     5149                                 74.0  ...           528.5   \n",
       "\n",
       "     dest_population  dest_density  DEP_DELAY_NEW ARR_DELAY_NEW CANCELLED  \\\n",
       "28           35151.0         528.5      13.302752     12.614679  0.000000   \n",
       "452        5046555.0        1419.9       8.532308      8.864198  0.006116   \n",
       "\n",
       "    Utilization  Regularity         ASM          RASM  \n",
       "28     0.643357    0.991701  222.893544  2.310071e+07  \n",
       "452    0.635778    0.991701  222.893544  2.310071e+07  \n",
       "\n",
       "[2 rows x 36 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for node in uncertain_nodes:\n",
    "    display(data[data['citypair'].str.contains(node)])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6982a154-7bfc-46fb-924c-1db359c297c8",
   "metadata": {},
   "source": [
    "### Link Prediction based on edge embeddings which are Hadamard product of node embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "8c575f0d-cae8-4ba3-856a-f50744ec01ae",
   "metadata": {},
   "outputs": [],
   "source": [
    "adj_matrix = np.array(nx.to_numpy_matrix(G))\n",
    "edge_embeddings_pos = []\n",
    "edge_embeddings_neg = []\n",
    "for node1_idx in range(adj_matrix.shape[0]):\n",
    "    for node2_idx in range(adj_matrix.shape[1]):\n",
    "        edge_embedding = np.multiply(node_embeddings[node1_idx], node_embeddings[node2_idx])\n",
    "        if adj_matrix[node1_idx, node2_idx] == 0:\n",
    "            edge_embeddings_neg.append(edge_embedding)\n",
    "        else:\n",
    "            edge_embeddings_pos.append(edge_embedding)\n",
    "edge_embeddings_pos = np.array(edge_embeddings_pos)\n",
    "edge_embeddings_neg = np.array(edge_embeddings_neg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "25fa7212-42f3-4331-ad69-29789ea1ac31",
   "metadata": {},
   "outputs": [],
   "source": [
    "X = np.concatenate([edge_embeddings_pos, edge_embeddings_neg], axis=0)\n",
    "y = np.concatenate([np.ones(len(edge_embeddings_pos)), np.zeros(len(edge_embeddings_neg))])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "49fd7f64-8fff-449a-aee5-17c1e8e540db",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "import numpy as np\n",
    "rng = np.random.RandomState(seed = 435)\n",
    "def bagging_fit(X, y, n_estimators, C = 1.0, max_iter = 100):\n",
    "    pos_samples = sum(y==1)\n",
    "    neg_samples = sum(y==0)\n",
    "    logistic_models = [LogisticRegression(C = C, max_iter = max_iter) for _ in range(n_estimators)]\n",
    "    for logistic_model in logistic_models:\n",
    "        bag_pos = np.random.choice(pos_samples, pos_samples, replace = True)\n",
    "        bag_neg = np.random.choice(neg_samples, neg_samples, replace = True)\n",
    "        bag = np.concatenate((bag_pos, bag_neg))\n",
    "        logistic_model.fit(X[bag,:], y[bag])\n",
    "    return logistic_models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "e98aa365-8fc3-45c1-a925-7d9030625f26",
   "metadata": {},
   "outputs": [],
   "source": [
    "def bagging_predict_neg_proba(X, models):\n",
    "    all_neg_edge_probs = np.array([logistic_model.predict_proba(X)[:,0] for logistic_model in models])\n",
    "    neg_edge_probs = np.mean(all_neg_edge_probs, axis = 0, keepdims = False)\n",
    "    return neg_edge_probs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "f32892ef-8eb4-4acd-875e-31b3f1228f58",
   "metadata": {},
   "outputs": [],
   "source": [
    "bag_ens = bagging_fit(X, y, n_estimators = 1000, C = 1, max_iter = 1000)\n",
    "neg_edge_probs = bagging_predict_neg_proba(X, bag_ens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "47b45afc-6c0d-4726-8ae1-08a6b21a50bc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Threshold = 0.301469, F-2Score = 1.210\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import precision_recall_curve\n",
    "precision, recall, thresholds = precision_recall_curve(y, (1-neg_edge_probs))\n",
    "\n",
    "# f-beta score\n",
    "beta = 2\n",
    "fscore = ((1 + beta)**2 * precision * recall) / (beta**2 * precision + recall + 1e-08)\n",
    "# locate the index of the largest f-beta score\n",
    "idx = np.argmax(fscore)\n",
    "best_f_beta = fscore[idx]\n",
    "best_thresh = thresholds[idx]\n",
    "print('Best Threshold = %f, F-2Score = %.3f' % (best_thresh, best_f_beta))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "be56c51e-18c8-4c6f-847a-aab4aa27b58d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(neg_edge_probs, bins = 50)\n",
    "plt.axvline(x = (1 - best_thresh), color = 'red')\n",
    "plt.text(0.75, 2, 'Best Threshold')\n",
    "plt.xlabel(\"Average Probability of Edges being Removed\")\n",
    "plt.ylabel(\"Frequency\")\n",
    "plt.title(\"Link Prediction (Stratified Bagging with Logistic Regression\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "11dea398-5c1a-43a4-b71b-65e9b7284d10",
   "metadata": {},
   "outputs": [],
   "source": [
    "removed_edges_idx = []\n",
    "num_nodes = len(G.nodes())\n",
    "for i in np.where(neg_edge_probs > (1-thresholds[idx]))[0]:\n",
    "    if adj_matrix[i//num_nodes][i%num_nodes] == 1:\n",
    "        removed_edges_idx.append((i//num_nodes, i%num_nodes))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "f21cd5b8-68b5-4202-9e35-e98a61cb94e1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of Edges being removed (logistic regression) is: 514\n"
     ]
    }
   ],
   "source": [
    "node_idx_pair = {idx: node for idx, node in enumerate(G.nodes())}\n",
    "removed_edges_lr = [(node_idx_pair[node1_idx], node_idx_pair[node2_idx])for node1_idx, node2_idx in removed_edges_idx]\n",
    "print(f\"The number of Edges being removed (logistic regression) is: {len(removed_edges_lr)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "ec61ad9a-f39b-4f65-9b72-bd938664c162",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "rf = RandomForestClassifier(n_estimators = 1000).fit(X,y)\n",
    "y_pred = rf.predict(X)\n",
    "removed_edges_idx = []\n",
    "num_nodes = len(G.nodes())\n",
    "for i in np.where(y_pred == 0)[0]:\n",
    "    if adj_matrix[i//num_nodes][i%num_nodes] == 1:\n",
    "        removed_edges_idx.append((i//num_nodes, i%num_nodes))\n",
    "removed_edges_rf = [(node_idx_pair[node1_idx], node_idx_pair[node2_idx])for node1_idx, node2_idx in removed_edges_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "c4160246-f77e-4416-809e-711c338f8de4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of Edges being removed (Random Forest) is: 712\n"
     ]
    }
   ],
   "source": [
    "print(f\"The number of Edges being removed (Random Forest) is: {len(removed_edges_rf)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "be203d39-942a-469d-8546-6ba3d8bd9208",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of Common Edges that can be removed is 455\n"
     ]
    }
   ],
   "source": [
    "print(f\"The number of Common Edges that can be removed is {len(set(removed_edges_lr).intersection(removed_edges_rf))}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "032b6347-0282-4967-a561-0300dc3072a9",
   "metadata": {},
   "outputs": [],
   "source": [
    "identified_hubs = ['BWI', 'MCO', 'MDW', 'AUS', 'DEN', 'SAT', 'PHX', 'LAS', 'PDX']\n",
    "removed_edges = set(removed_edges_rf).intersection(removed_edges_lr)\n",
    "removed_edges_wo_hubs = []\n",
    "for node1, node2 in removed_edges:\n",
    "    if (not node1 in identified_hubs) or (not node1 in identified_hubs):\n",
    "        removed_edges_wo_hubs.append((node1, node2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "5709189d-941e-4c85-8d0c-6cb761b9c673",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The number of Edges that don't contain potential hubs being removed is: 423\n"
     ]
    }
   ],
   "source": [
    "print(f\"The number of Edges that don't contain potential hubs being removed is: {len(removed_edges_wo_hubs)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "601bc6f7-45f6-4ad6-8717-bb2285404b69",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{('AUS', 'DAL'),\n",
       " ('AUS', 'DCA'),\n",
       " ('AUS', 'ELP'),\n",
       " ('AUS', 'HRL'),\n",
       " ('AUS', 'MSY'),\n",
       " ('AUS', 'SAN'),\n",
       " ('PDX', 'BUR'),\n",
       " ('PDX', 'DAL'),\n",
       " ('PDX', 'DEN'),\n",
       " ('PDX', 'LAS'),\n",
       " ('PDX', 'LAX'),\n",
       " ('PDX', 'MDW'),\n",
       " ('PDX', 'OAK'),\n",
       " ('PDX', 'ONT'),\n",
       " ('PDX', 'STL'),\n",
       " ('PHX', 'MCI'),\n",
       " ('PHX', 'SMF'),\n",
       " ('SAT', 'BWI'),\n",
       " ('SAT', 'DAL'),\n",
       " ('SAT', 'DEN'),\n",
       " ('SAT', 'FLL'),\n",
       " ('SAT', 'HOU'),\n",
       " ('SAT', 'LAS'),\n",
       " ('SAT', 'LAX'),\n",
       " ('SAT', 'MCI'),\n",
       " ('SAT', 'MCO'),\n",
       " ('SAT', 'MDW'),\n",
       " ('SAT', 'OAK'),\n",
       " ('SAT', 'PHX'),\n",
       " ('SAT', 'SAN'),\n",
       " ('SAT', 'STL'),\n",
       " ('SAT', 'TPA')}"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "set(removed_edges).difference(removed_edges_wo_hubs)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}