twitter_mentions_network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import networkx as nx\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as pl\n",
    "from collections import defaultdict\n",
    "from pprint import pprint\n",
    "import itertools\n",
    "import random\n",
    "\n",
    "pl.rcParams['figure.figsize'] = [12, 12]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Name: \n",
      "Type: Graph\n",
      "Number of nodes: 7175\n",
      "Number of edges: 11681\n",
      "Average degree:   3.2560\n",
      "{'cen': 3271, 'con': 2101, 'lib': 1803, 'total': 7175}\n"
     ]
    }
   ],
   "source": [
    "g = nx.read_graphml('mention.graphml')\n",
    "if g.is_directed():\n",
    "    g = g.to_undirected()\n",
    "# convert from multigraph to standard graph\n",
    "g = nx.Graph(g)\n",
    "nodes = g.nodes(data=True)\n",
    "edges = g.edges()\n",
    "info = nx.info(g)\n",
    "print(info)\n",
    "nodes = {k: v for k, v in nodes}\n",
    "edges = tuple((k, v) for k, v in edges)\n",
    "fname = 'cluster'\n",
    "fcon = 'right'\n",
    "flib = 'left'\n",
    "fcen = '-'\n",
    "con = [k for k, v in nodes.items() if v[fname] == fcon]\n",
    "lib = [k for k, v in nodes.items() if v[fname] == flib]\n",
    "cen = [k for k, v in nodes.items() if v[fname] == fcen]\n",
    "pprint(dict(con=len(con), lib=len(lib), cen=len(cen), total=(len(con) + len(lib) + len(cen))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'modularity': 0.10937366167058862}\n"
     ]
    }
   ],
   "source": [
    "# calc modularity of network\n",
    "from networkx.algorithms.community.quality import modularity\n",
    "q =  modularity(g, [con, lib, cen])\n",
    "\n",
    "pprint(dict(modularity=q))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Edge Sample:\n",
      "[('n2677', 'n4345'),\n",
      " ('n5785', 'n6827'),\n",
      " ('n1111', 'n1960'),\n",
      " ('n714', 'n1145'),\n",
      " ('n3471', 'n5937'),\n",
      " ('n1704', 'n4648'),\n",
      " ('n2201', 'n4477'),\n",
      " ('n1199', 'n3913'),\n",
      " ('n3534', 'n4082'),\n",
      " ('n4023', 'n4900'),\n",
      " ('n234', 'n3704'),\n",
      " ('n5018', 'n5783'),\n",
      " ('n714', 'n5921'),\n",
      " ('n92', 'n3198'),\n",
      " ('n551', 'n568'),\n",
      " ('n1218', 'n6259'),\n",
      " ('n1882', 'n4193'),\n",
      " ('n802', 'n6447'),\n",
      " ('n1060', 'n3152'),\n",
      " ('n3219', 'n6667'),\n",
      " ('n2460', 'n4240'),\n",
      " ('n4570', 'n5978'),\n",
      " ('n3720', 'n6278'),\n",
      " ('n1882', 'n4814'),\n",
      " ('n1267', 'n3164')]\n",
      "Nodes from Edges Sample:\n",
      "{'-': {'n234': (...,),\n",
      "       'n2677': (...,),\n",
      "       'n3152': (...),\n",
      "       'n3219': (...,),\n",
      "       'n4193': (...,),\n",
      "       'n4814': (...),\n",
      "       'n5937': (...),\n",
      "       'n6278': (...)},\n",
      " 'left': {'n1060': (...),\n",
      "          'n3164': (...),\n",
      "          'n3534': (...,),\n",
      "          'n3704': (...),\n",
      "          'n3913': (...),\n",
      "          'n4082': (...),\n",
      "          'n4240': (...),\n",
      "          'n4345': (...),\n",
      "          'n4477': (...),\n",
      "          'n4648': (...),\n",
      "          'n4900': (...),\n",
      "          'n5018': (...),\n",
      "          'n568': (...),\n",
      "          'n5783': (...),\n",
      "          'n6827': (...)},\n",
      " 'right': {'n1111': (...),\n",
      "           'n1145': (...),\n",
      "           'n1199': (...),\n",
      "           'n1218': (...),\n",
      "           'n1267': (...,),\n",
      "           'n1704': (...),\n",
      "           'n1882': (...),\n",
      "           'n1960': (...),\n",
      "           'n2201': (...),\n",
      "           'n2460': (...),\n",
      "           'n3198': (...),\n",
      "           'n3471': (...),\n",
      "           'n3720': (...),\n",
      "           'n4023': (...),\n",
      "           'n4570': (...),\n",
      "           'n551': (...),\n",
      "           'n5785': (...),\n",
      "           'n5921': (...),\n",
      "           'n5978': (...,),\n",
      "           'n6259': (...),\n",
      "           'n6447': (...),\n",
      "           'n6667': (...),\n",
      "           'n714': (...),\n",
      "           'n802': (...),\n",
      "           'n92': (...,)}}\n"
     ]
    }
   ],
   "source": [
    "esample = random.sample(edges, 25)\n",
    "print(\"Edge Sample:\")\n",
    "pprint(esample)\n",
    "\n",
    "nsample = {'left': {}, 'right': {}, '-': {}}\n",
    "for u, v in esample:\n",
    "    uclust = g.nodes[u]['cluster']\n",
    "    vclust = g.nodes[v]['cluster']\n",
    "    if u not in nsample[uclust]:\n",
    "        nsample[uclust][u] = tuple(g.neighbors(u))\n",
    "    if v not in nsample[vclust]:\n",
    "        nsample[vclust][v] = tuple(g.neighbors(v))\n",
    "\n",
    "print(\"Nodes from Edges Sample:\")\n",
    "pprint(nsample, depth=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n3189 1\n"
     ]
    }
   ],
   "source": [
    "def ratio_calc(graph, node, sweight=1, oweight=0, nweight=0.5, \n",
    "               field='cluster', neutral='-', right='right', left='left', \n",
    "               no_alignment_value=None):\n",
    "    \"\"\"\n",
    "    Perform a ratio calculation on given node in graph.\n",
    "    \"\"\"\n",
    "    nbors = tuple(graph.neighbors(node))\n",
    "    nclust = graph.nodes[node][field]\n",
    "    \n",
    "    if nclust == neutral:\n",
    "        return no_alignment_value\n",
    "    \n",
    "    counts = { k: float(len(list(v))) for k,v in itertools.groupby(nbors, key=lambda y: graph.nodes[y][field]) }\n",
    "    nother = right if nclust == left else left\n",
    "    ratio = ((sweight * counts.get(nclust, 0)) + (oweight * counts.get(nother, 0)) + (nweight * counts.get(neutral, 0))) \\\n",
    "                / len(nbors)\n",
    "    return ratio\n",
    "\n",
    "ratios = { n: ratio_calc(g, n) for n in g.nodes() }\n",
    "ratios = { k: v for k, v in ratios.items() if v }\n",
    "#pprint(ratios)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x864 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "d = { k: len(tuple(v)) for k, v in itertools.groupby(sorted(ratios.values())) }\n",
    "plot, (ap, bp) = pl.subplots(2)\n",
    "plot.suptitle('Mention Ratio Distribution')\n",
    "ap.scatter(d.keys(), d.values())\n",
    "bp.hist(ratios.values(), bins=9)\n",
    "plot.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}