expert_rank_public.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "78b94a69-678a-4a3a-8ba0-55f375232447",
   "metadata": {},
   "source": [
    "This is code example for my article on Medium. \n",
    "Fulltext is available here:\n",
    "\n",
    "https://medium.com/exness-blog/from-employee-voices-to-expert-stars-uncovering-the-most-valuable-assets-in-organizations-7ae0654df092\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "091e3d33-3a33-45a6-93c1-01b99168828e",
   "metadata": {},
   "source": [
    "# Experts Rank project\n",
    "\n",
    "### General idea: \n",
    "\n",
    "We have several levels of employees: \n",
    "* experts, \n",
    "* managers, \n",
    "* top-manegers (C-levels). \n",
    "\n",
    "We want to know which experts are the most respectful among all the other categories of employees. To find out the answer we ask all the employees which experts they respect the most. \n",
    "\n",
    "### Approach\n",
    "\n",
    "Answer of every employee looks like that: \n",
    "```sql \n",
    "'user_1': ['user_2', 'user_3', 'user_19']\n",
    "```\n",
    "\n",
    "where \n",
    "* user_1 - employee who filled the questionnaire. \n",
    "* user_2, user_3, user_19 - experts, mentioned by user_1\n",
    "\n",
    "### Suggested algorithm\n",
    "\n",
    "We can use Pagerank algorithm to analyze all the amount of answers.\n",
    "\n",
    "The algorithm of PageRank works by counting the number and quality of links to an expert to determine a rough estimate of how important the expert is. The underlying assumption is that more important experts are likely to receive more links from other employees and then could partially transfer their value to the experts they choose.\n",
    "\n",
    "<div>\n",
    "<img src=\"img/pagerank.png\" width=\"700\"/>\n",
    "</div>\n",
    "\n",
    "Experts as nodes and mentiones as edges, taking into consideration 'authority hubs'. The rank value indicates the importance of a particular expert. A link to another employee counts as a vote of support. The PageRank of a page is defined recursively and depends on the number and PageRank metric of all pages that link to it. An employee that is linked to by employee with high PageRank receives a higher rank itself (figure 'C')."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d0b0fbfd-f070-4aed-a0ec-935d484dc2c0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'1.10.0'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import networkx as nx\n",
    "\n",
    "np.set_printoptions(suppress=True,\n",
    "   formatter={'float_kind':'{:0.3f}'.format})  # float, 3 units precision\n",
    "\n",
    "import scipy\n",
    "scipy.__version__  # must be higher than 1.8.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "016d0f9a-7e1c-4e1a-a647-b1f8898de1ea",
   "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>source</th>\n",
       "      <th>target</th>\n",
       "      <th>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>user_1</td>\n",
       "      <td>user_2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>user_1</td>\n",
       "      <td>user_3</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>user_1</td>\n",
       "      <td>user_4</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>user_2</td>\n",
       "      <td>user_1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>user_2</td>\n",
       "      <td>user_3</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>user_3</td>\n",
       "      <td>user_2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>user_4</td>\n",
       "      <td>user_2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>user_4</td>\n",
       "      <td>user_3</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   source  target  weight\n",
       "0  user_1  user_2       1\n",
       "1  user_1  user_3       1\n",
       "2  user_1  user_4       1\n",
       "3  user_2  user_1       1\n",
       "4  user_2  user_3       1\n",
       "5  user_3  user_2       1\n",
       "6  user_4  user_2       1\n",
       "7  user_4  user_3       1"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Nodes could be digits\n",
    "edges = pd.DataFrame({\n",
    "        \"source\": [1, 1, 1, 2, 2, 3, 4, 4],\n",
    "        \"target\": [2, 3, 4, 1, 3, 2, 2, 3],\n",
    "        \"weight\": [1, 1, 1, 1, 1, 1, 1, 1],})\n",
    "\n",
    "# Or text\n",
    "edges = pd.DataFrame({\n",
    "        \"source\": ['user_1', 'user_1', 'user_1', 'user_2', \n",
    "                   'user_2', 'user_3', 'user_4', 'user_4'],\n",
    "        \"target\": ['user_2', 'user_3', 'user_4', 'user_1', \n",
    "                   'user_3', 'user_2', 'user_2', 'user_3'],\n",
    "        \"weight\": [1, 1, 1, 1, 1, 1, 1, 1],})\n",
    "\n",
    "edges"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "ba200059-fae2-4430-befe-3dee1153d108",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Building Directed Graph with self loops: \n",
    "\n",
    "G = nx.from_pandas_edgelist(edges, \n",
    "                            edge_attr=True, \n",
    "                            create_using=nx.DiGraph)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "0464095e-e7b4-4fd2-a3e2-336c62adb6e0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'user_2': {'weight': 1}, 'user_3': {'weight': 1}, 'user_4': {'weight': 1}}\n",
      "{'user_1': {'weight': 1}, 'user_3': {'weight': 1}}\n",
      "{'user_2': {'weight': 1}}\n",
      "{'user_2': {'weight': 1}, 'user_3': {'weight': 1}}\n"
     ]
    }
   ],
   "source": [
    "for i in G: \n",
    "    print(G[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2f4f6de1-7b61-496b-89aa-d3e1d112c989",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "user_1 0.205\n",
      "user_2 0.395\n",
      "user_3 0.304\n",
      "user_4 0.096\n"
     ]
    }
   ],
   "source": [
    "pr = nx.pagerank(G, \n",
    "                 alpha=0.85, \n",
    "                 max_iter=100)\n",
    "\n",
    "for i in pr:\n",
    "    print(i, round(pr[i], 3))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2b319c46-94eb-47c8-bdd9-1c731b80b4b0",
   "metadata": {},
   "source": [
    "Let's check ourselves with Page Rank Simulator: \n",
    "https://computerscience.chemeketa.edu/cs160Reader/_static/pageRankApp/index.html\n",
    "\n",
    "\n",
    "<div>\n",
    "<img src=\"img/toy_ex.png\" width=\"700\"/>\n",
    "</div>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "db6baf0d-d2f8-41cf-92e8-97bd7054ed57",
   "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>user_1</th>\n",
       "      <th>user_2</th>\n",
       "      <th>user_3</th>\n",
       "      <th>user_4</th>\n",
       "      <th>user_5</th>\n",
       "      <th>user_6</th>\n",
       "      <th>user_7</th>\n",
       "      <th>user_8</th>\n",
       "      <th>user_9</th>\n",
       "      <th>user_10</th>\n",
       "      <th>user_11</th>\n",
       "      <th>user_12</th>\n",
       "      <th>user_13</th>\n",
       "      <th>user_14</th>\n",
       "      <th>user_15</th>\n",
       "      <th>user_16</th>\n",
       "      <th>user_17</th>\n",
       "      <th>user_18</th>\n",
       "      <th>user_19</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>user_2</td>\n",
       "      <td>user_1</td>\n",
       "      <td>user_1</td>\n",
       "      <td>user_1</td>\n",
       "      <td>user_2</td>\n",
       "      <td>user_1</td>\n",
       "      <td>user_1</td>\n",
       "      <td>user_5</td>\n",
       "      <td>user_5</td>\n",
       "      <td>user_3</td>\n",
       "      <td>user_10</td>\n",
       "      <td>user_10</td>\n",
       "      <td>user_10</td>\n",
       "      <td>user_10</td>\n",
       "      <td>user_4</td>\n",
       "      <td>user_15</td>\n",
       "      <td>user_15</td>\n",
       "      <td>user_15</td>\n",
       "      <td>user_18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>user_3</td>\n",
       "      <td>user_5</td>\n",
       "      <td>user_14</td>\n",
       "      <td>user_19</td>\n",
       "      <td>user_6</td>\n",
       "      <td>user_9</td>\n",
       "      <td>user_9</td>\n",
       "      <td>None</td>\n",
       "      <td>user_6</td>\n",
       "      <td>user_11</td>\n",
       "      <td>user_12</td>\n",
       "      <td>user_14</td>\n",
       "      <td>user_14</td>\n",
       "      <td>user_9</td>\n",
       "      <td>user_19</td>\n",
       "      <td>user_17</td>\n",
       "      <td>user_16</td>\n",
       "      <td>None</td>\n",
       "      <td>user_17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>user_19</td>\n",
       "      <td>user_7</td>\n",
       "      <td>user_13</td>\n",
       "      <td>user_15</td>\n",
       "      <td>user_7</td>\n",
       "      <td>user_2</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>user_9</td>\n",
       "      <td>user_3</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>user_9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>user_1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    user_1  user_2   user_3   user_4  user_5  user_6  user_7  user_8  user_9  \\\n",
       "0   user_2  user_1   user_1   user_1  user_2  user_1  user_1  user_5  user_5   \n",
       "1   user_3  user_5  user_14  user_19  user_6  user_9  user_9    None  user_6   \n",
       "2  user_19  user_7  user_13  user_15  user_7  user_2    None    None    None   \n",
       "3     None    None     None     None    None    None    None    None    None   \n",
       "\n",
       "   user_10  user_11  user_12  user_13  user_14  user_15  user_16  user_17  \\\n",
       "0   user_3  user_10  user_10  user_10  user_10   user_4  user_15  user_15   \n",
       "1  user_11  user_12  user_14  user_14   user_9  user_19  user_17  user_16   \n",
       "2     None   user_9   user_3     None     None     None     None     None   \n",
       "3     None     None     None     None     None     None     None     None   \n",
       "\n",
       "   user_18  user_19  \n",
       "0  user_15  user_18  \n",
       "1     None  user_17  \n",
       "2     None   user_9  \n",
       "3     None   user_1  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Let's check several cases:\n",
    "\n",
    "d = {'user_1': ['user_2', 'user_3', 'user_19',  None], # Imagine this is CEO, \n",
    "                                                       # he mention (links) 2 his C-levels (user_2, user_3) and 1 expert\n",
    "     \n",
    "      # C-levels\n",
    "     'user_2': ['user_1', 'user_5', 'user_7',   None], # C-level 1, links CEO and 2 his his managers down below\n",
    "     'user_3': ['user_1', 'user_14','user_13',  None], # C-level 2, links CEO and 2 his his managers down below\n",
    "     'user_4': ['user_1', 'user_19','user_15',  None], # C-level 3, links CEO and 2 his his managers down below\n",
    "\n",
    "     \n",
    "      # Department of C-level 1\n",
    "     'user_5': ['user_2', 'user_6',  'user_7',  None], # Manager of C-level 1, links his 'C-level 1' and 2 experts\n",
    "     'user_6': ['user_1', 'user_9',  'user_2',  None], # Expert  of C-level 1, links CEO and 1 expert\n",
    "     'user_7': ['user_1', 'user_9',  None,      None], # Expert  of C-level 1, links CEO and 1 expert\n",
    "     'user_8': ['user_5', None,      None,      None], # Expert  of C-level 1, links his manager and 1 expert  \n",
    "                                                       # === Never mentioned user\n",
    "     'user_9': ['user_5', 'user_6',  None,      None], # Expert  of C-level 1, links his manager and 1 expert\n",
    "                                                       # === Hidden expert, C-levels never mentioned him\n",
    "\n",
    "      # Department of C-level 2\n",
    "     'user_10': ['user_3', 'user_11',  None,    None], # Manager of C-level 2, links his 'C-level 1' and 1 expert\n",
    "     'user_11': ['user_10','user_12', 'user_9', None], # Expert  of C-level 2, links his manager and 1 expert\n",
    "     'user_12': ['user_10','user_14', 'user_3', None], # Expert  of C-level 2, links his manager and 1 expert\n",
    "     'user_13': ['user_10','user_14',  None,    None], # Expert  of C-level 2, links his manager and 1 expert\n",
    "     'user_14': ['user_10', 'user_9',  None,    None], # Expert  of C-level 2, links (!) Expert of another Dep\n",
    "     \n",
    "      # Department of C-level 3\n",
    "     'user_15': ['user_4','user_19',   None,    None], # Manager of C-level 3, links his 'C-level 1' and 1 expert\n",
    "     'user_16': ['user_15','user_17',  None,    None], # Casual employee of C-level 3, links his manager and 1 expert\n",
    "     'user_17': ['user_15','user_16',  None,    None], # Casual employee of C-level 3, links his manager and 1 expert\n",
    "     'user_18': ['user_15', None ,     None,    None], # Casual employee of C-level 3, links his manager and 1 expert\n",
    "     'user_19': ['user_18', 'user_17',\n",
    "                                   'user_9','user_1']  # Expert  of C-level 3, link to CEO and linked by CEO \n",
    "                                                       # === Well-known expert case\n",
    "    }\n",
    "\n",
    "df = pd.DataFrame(data=d)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "0f25511b-b05d-41c6-99a7-abb77c2eeacc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "amount of edges:  46\n"
     ]
    },
    {
     "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>source</th>\n",
       "      <th>target</th>\n",
       "      <th>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>user_1</td>\n",
       "      <td>user_2</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>user_10</td>\n",
       "      <td>user_3</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>user_11</td>\n",
       "      <td>user_10</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>user_12</td>\n",
       "      <td>user_10</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>user_13</td>\n",
       "      <td>user_10</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    source   target  weight\n",
       "0   user_1   user_2       1\n",
       "1  user_10   user_3       1\n",
       "2  user_11  user_10       1\n",
       "3  user_12  user_10       1\n",
       "4  user_13  user_10       1"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_source = df.T.sort_index()\n",
    "df_source = df_source.reset_index()\n",
    "df_source = df_source.rename(columns={0: \"target\"})\n",
    "\n",
    "df_edges_1 = df_source[['index', 1]]\n",
    "df_edges_1 = df_edges_1.rename(columns={1: \"target\"})\n",
    "\n",
    "df_edges_2 = df_source[['index', 2]]\n",
    "df_edges_2 = df_edges_2.rename(columns={2: \"target\"})\n",
    "\n",
    "df_edges_3 = df_source[['index', 3]]\n",
    "df_edges_3 = df_edges_3.rename(columns={3: \"target\"})\n",
    "\n",
    "df_unioned = df_source[['index', 'target']].append(df_edges_1, ignore_index=True)\n",
    "df_unioned = df_unioned.append(df_edges_2, ignore_index=True)\n",
    "df_unioned = df_unioned.append(df_edges_3, ignore_index=True)\n",
    "\n",
    "df_unioned = df_unioned[df_unioned['target'].notna()]\n",
    "df_unioned['weight'] = 1\n",
    "df_unioned = df_unioned.rename(columns={'index':'source'})\n",
    "\n",
    "print('amount of edges: ', len(df_unioned))\n",
    "\n",
    "df_unioned.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "7c82bb65-1109-41c0-ba21-536f3a9e1ad7",
   "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>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>user_1</th>\n",
       "      <td>0.120170</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_9</th>\n",
       "      <td>0.094091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_2</th>\n",
       "      <td>0.084014</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_5</th>\n",
       "      <td>0.078397</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_19</th>\n",
       "      <td>0.078248</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_6</th>\n",
       "      <td>0.070095</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_3</th>\n",
       "      <td>0.068164</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_15</th>\n",
       "      <td>0.062460</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_7</th>\n",
       "      <td>0.053910</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_10</th>\n",
       "      <td>0.050861</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_14</th>\n",
       "      <td>0.043377</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_4</th>\n",
       "      <td>0.034441</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_17</th>\n",
       "      <td>0.034024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_11</th>\n",
       "      <td>0.029511</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_13</th>\n",
       "      <td>0.027208</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_18</th>\n",
       "      <td>0.024523</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_16</th>\n",
       "      <td>0.022355</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_12</th>\n",
       "      <td>0.016256</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>user_8</th>\n",
       "      <td>0.007895</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           weight\n",
       "user_1   0.120170\n",
       "user_9   0.094091\n",
       "user_2   0.084014\n",
       "user_5   0.078397\n",
       "user_19  0.078248\n",
       "user_6   0.070095\n",
       "user_3   0.068164\n",
       "user_15  0.062460\n",
       "user_7   0.053910\n",
       "user_10  0.050861\n",
       "user_14  0.043377\n",
       "user_4   0.034441\n",
       "user_17  0.034024\n",
       "user_11  0.029511\n",
       "user_13  0.027208\n",
       "user_18  0.024523\n",
       "user_16  0.022355\n",
       "user_12  0.016256\n",
       "user_8   0.007895"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "G = nx.from_pandas_edgelist(df_unioned, source='source', \n",
    "                            target='target', \n",
    "                            edge_attr=True, create_using=nx.DiGraph)\n",
    "\n",
    "pr = nx.pagerank(G, alpha=0.85, max_iter=100) \n",
    "\n",
    "df_pr = pd.DataFrame(data=pr, index=[0]).T\n",
    "df_pr = df_pr.rename(columns={0:'weight'})\n",
    "df_pr = df_pr.sort_values(by=['weight'], ascending=False)\n",
    "\n",
    "df_pr.head(40)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "029f2ba0-09e6-42f9-b124-ebce586475da",
   "metadata": {},
   "source": [
    "# Next steps: \n",
    "\n",
    "We can upload result of that algo to external tool like Gephi to create simple map in 2-dimentional space: \n",
    "\n",
    "<div>\n",
    "<img src=\"img/gephi_viz.png\" width=\"850\"/>\n",
    "</div>\n",
    "\n",
    "Thick of line and size of dot is based on rank of expert."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "74ecac12-08a2-4b0d-8abf-635dfce29560",
   "metadata": {},
   "source": [
    "### Useful links\n",
    "\n",
    "* PageRank on wiki: \n",
    "https://en.wikipedia.org/wiki/PageRank\n",
    "\n",
    "* PageRank simulator: \n",
    "https://computerscience.chemeketa.edu/cs160Reader/_static/pageRankApp/index.html\n",
    "\n",
    "* Identifying Potential Managerial Personnel Using PageRank and Social Network Analysis: The Case Study of a European IT Company:\n",
    "https://www.mdpi.com/2076-3417/11/15/6985\n",
    "\n",
    "* Professional Evaluation. Application of the PageRank Algorithm in Employee Rating: http://archive.sciendo.com/EAM/eam.2014.15.issue-2/eam-2014-0015/eam-2014-0015.pdf\n",
    "\n",
    "* Description of project idea: \n",
    "https://medium.com/exness-blog/from-employee-voices-to-expert-stars-uncovering-the-most-valuable-assets-in-organizations-7ae0654df092\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f2488330-0564-4c3f-8527-c31098045e56",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.9.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}