Add initial twitter results plots

investigate-results.ipynb

@@ -11,7 +11,9 @@
     "2. Plot **ROC curves** for inputted network, visibility, and link prediction method\n",
     "3. **Save ROC AUC/AP plots** as .pdf files\n",
     "4. **Save minimum runtime plots** as .pdf files\n",
-    "5. **Save network statistics** as .txt files"
+    "5. **Save network statistics** as .txt files\n",
+    "6. Generate **ROC Plots by Algorithm**\n",
+    "7. Generate **Twitter Results** Plots"
    ]
   },
   {
@@ -548,7 +550,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 76,
+   "execution_count": 14,
    "metadata": {
     "collapsed": true
    },
@@ -696,7 +698,9 @@
   {
    "cell_type": "code",
    "execution_count": 271,
-   "metadata": {},
+   "metadata": {
+    "collapsed": true
+   },
    "outputs": [],
    "source": [
     "# Generate AUROC graphs for each algorithm, frac_hidden\n",
@@ -712,6 +716,139 @@
     "                                  sizes='small')"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "collapsed": true
+   },
+   "source": [
+    "## 7. Twitter Results Plots"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Runtime Plots"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import json\n",
+    "RESULTS_FOLDER = \"./results/\"\n",
+    "\n",
+    "# Get minimum runtimes across Twitter experiments\n",
+    "twitter_min_runtimes = {}\n",
+    "TWITTER_EXPERIMENT_NUMS = [0, 1, 2]\n",
+    "\n",
+    "# Iterate over all Twitter experiments, store minimum runtimes\n",
+    "for twitter_num in TWITTER_EXPERIMENT_NUMS:\n",
+    "    # Load in results for this experiment\n",
+    "    curr_fb_dict = None\n",
+    "    twitter_filename = './results/twitter-experiment-{}-results.json'.format(twitter_num)\n",
+    "    with open(twitter_filename, 'rb') as f:\n",
+    "        curr_twitter_dict = json.load(f)\n",
+    "        \n",
+    "    # Iterate over experiments in curr_twitter_dict\n",
+    "    for experiment, results in curr_twitter_dict.iteritems():\n",
+    "        # Create (method --> runtime) dict, if necessary\n",
+    "        if experiment not in twitter_min_runtimes:\n",
+    "            twitter_min_runtimes[experiment] = {}\n",
+    "        \n",
+    "        # Iterate over link prediction methods in current network & experiment\n",
+    "        for method, metrics in results.iteritems():\n",
+    "            # Method currently has recorded runtime\n",
+    "            if method in twitter_min_runtimes[experiment]: \n",
+    "                # Overwrite previous runtime\n",
+    "                if metrics['runtime'] < twitter_min_runtimes[experiment][method]:\n",
+    "                    twitter_min_runtimes[experiment][method] = metrics['runtime']\n",
+    "            # Method does not yet have recorded runtime\n",
+    "            else:\n",
+    "                twitter_min_runtimes[experiment][method] = metrics['runtime']"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Plot runtime bar plots for all experiments\n",
+    "import matplotlib.pyplot as plt\n",
+    "for experiment_name, runtime_dict in twitter_min_runtimes.iteritems():\n",
+    "    descriptors = experiment_name.split('-')\n",
+    "    ego_num = descriptors[1]\n",
+    "    amt_hidden = descriptors[2]\n",
+    "    graph_title = \"Link Prediction Runtimes: Twitter-{} graph, {} hidden\".format(ego_num, amt_hidden)\n",
+    "    \n",
+    "    plt.figure(figsize=(8, 4))\n",
+    "    plt.bar(range(len(runtime_dict)), runtime_dict.values())\n",
+    "    plt.xticks(range(len(runtime_dict)), list(runtime_dict.keys()))\n",
+    "    plt.grid(linestyle='dashed')\n",
+    "    \n",
+    "    plt.xlabel(\"Link prediction methods\")\n",
+    "    plt.ylabel(\"Minimum runtime (seconds)\")\n",
+    "    plt.title(graph_title)\n",
+    "    \n",
+    "    plt.savefig('./result-plots-by-graph/' + experiment_name + '-runtime.pdf')\n",
+    "    plt.clf()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### ROC AUC Plots"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Save bar plot in .pdf given metric name, metric dict (with results for each method), experiment name\n",
+    "RESULT_PLOTS_FOLDER = './result-plots-by-graph/'\n",
+    "twitter_results = None\n",
+    "with open('./results/twitter-experiment-0-results.json', 'rb') as f:\n",
+    "    twitter_results = json.load(f)\n",
+    "    \n",
+    "# Save Twitter results bar plot\n",
+    "def save_twitter_bar_plot(metric_name, metric_dict, experiment_name):\n",
+    "    descriptors = experiment_name.split('-')\n",
+    "    ego_num = descriptors[1]\n",
+    "    amt_hidden = descriptors[2]\n",
+    "    graph_title = \"Link Prediction Results: Twitter-{} graph, {} hidden\".format(ego_num, amt_hidden)\n",
+    "    \n",
+    "    plt.figure(figsize=(8, 4))\n",
+    "    plt.bar(range(len(metric_dict)), metric_dict.values())\n",
+    "    plt.xticks(range(len(metric_dict)), list(metric_dict.keys()))\n",
+    "    plt.grid(linestyle='dashed')\n",
+    "    \n",
+    "    plt.xlabel(\"Link prediction methods\")\n",
+    "    plt.ylabel(\"Metric: {}\".format(metric_name))\n",
+    "    plt.title(graph_title)\n",
+    "    \n",
+    "    plt.savefig(RESULT_PLOTS_FOLDER + experiment_name + '-' + metric_name + '.pdf')\n",
+    "    plt.clf()\n",
+    "    \n",
+    "# Plot ROC AUC bar graphs for each amt_hidden\n",
+    "for experiment_name, results_dict in twitter_results.iteritems():\n",
+    "    roc_dict = dict()\n",
+    "        \n",
+    "    # Populate metric_dict with results for each method\n",
+    "    for method, method_results_dict in results_dict.iteritems():\n",
+    "        roc_dict[method] = method_results_dict['test_roc']\n",
+    "\n",
+    "    # Generate plot for this method and experiment name\n",
+    "    save_twitter_bar_plot(\"test_roc\", roc_dict, experiment_name)"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,