Graph_Clustering_using_GraphSAGE.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Graph_Clustering_using_GraphSAGE.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "68aV74HDsfp8"
      },
      "source": [
        "#### Importing Libraries"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "snG9DmkG_pOD",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0b4e635f-7033-4255-a499-120da3771851"
      },
      "source": [
        "import sys\n",
        "if 'google.colab' in sys.modules:\n",
        "  %pip install -q stellargraph[demos]==1.0.0rc1\n",
        "\n",
        "  # verify that we're using the correct version of StellarGraph for this notebook\n",
        "import stellargraph as sg\n",
        "\n",
        "try:\n",
        "    sg.utils.validate_notebook_version(\"1.0.0rc1\")\n",
        "except AttributeError:\n",
        "    raise ValueError(\n",
        "        f\"This notebook requires StellarGraph version 1.0.0rc1, but a different version {sg.__version__} is installed.  Please see <https://github.com/stellargraph/stellargraph/issues/1172>.\"\n",
        "    ) from None\n"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[K     |████████████████████████████████| 378kB 4.0MB/s \n",
            "\u001b[K     |████████████████████████████████| 235kB 22.3MB/s \n",
            "\u001b[K     |████████████████████████████████| 51kB 5.8MB/s \n",
            "\u001b[?25h  Building wheel for mplleaflet (setup.py) ... \u001b[?25l\u001b[?25hdone\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VQPotB4L_77z"
      },
      "source": [
        "import networkx as nx\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import os\n",
        "import random\n",
        "\n",
        "import stellargraph as sg\n",
        "from stellargraph.data import EdgeSplitter\n",
        "from stellargraph.mapper import GraphSAGELinkGenerator\n",
        "from stellargraph.layer import GraphSAGE, link_classification\n",
        "from stellargraph.data import UniformRandomWalk\n",
        "from stellargraph.data import UnsupervisedSampler\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "from tensorflow import keras\n",
        "from sklearn import preprocessing, feature_extraction, model_selection\n",
        "from sklearn.linear_model import LogisticRegressionCV, LogisticRegression\n",
        "from sklearn.metrics import accuracy_score\n",
        "\n",
        "from stellargraph import globalvar\n",
        "\n",
        "from stellargraph import datasets\n",
        "from IPython.display import display, HTML"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FPu-fu2dsm5w"
      },
      "source": [
        "#### Data Loading"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qxVmdAO9_9MJ",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 424
        },
        "outputId": "1a927d59-d556-4716-cf54-cf6b2d52db45"
      },
      "source": [
        "keyword_links=pd.read_csv('/content/stack_network_links.csv')\n",
        "keyword_links=keyword_links.drop(columns=\"value\")\n",
        "keyword_links"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "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",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>azure</td>\n",
              "      <td>.net</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>sql-server</td>\n",
              "      <td>.net</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>asp.net</td>\n",
              "      <td>.net</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>entity-framework</td>\n",
              "      <td>.net</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>wpf</td>\n",
              "      <td>.net</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>485</th>\n",
              "      <td>objective-c</td>\n",
              "      <td>xcode</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>486</th>\n",
              "      <td>swift</td>\n",
              "      <td>xcode</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>487</th>\n",
              "      <td>iphone</td>\n",
              "      <td>xcode</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>488</th>\n",
              "      <td>ios</td>\n",
              "      <td>xcode</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>489</th>\n",
              "      <td>json</td>\n",
              "      <td>xml</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>490 rows × 2 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "               source target\n",
              "0               azure   .net\n",
              "1          sql-server   .net\n",
              "2             asp.net   .net\n",
              "3    entity-framework   .net\n",
              "4                 wpf   .net\n",
              "..                ...    ...\n",
              "485       objective-c  xcode\n",
              "486             swift  xcode\n",
              "487            iphone  xcode\n",
              "488               ios  xcode\n",
              "489              json    xml\n",
              "\n",
              "[490 rows x 2 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IC0fuiWwBqun",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 455
        },
        "outputId": "75bb44ba-c561-4a12-ebef-0b88d68e99e2"
      },
      "source": [
        "keyword_features=pd.read_csv('/content/stack_network_nodes.csv')\n",
        "keyword_features=keyword_features.drop(columns=\"nodesize\")\n",
        "keyword_features=keyword_features.set_index(\"name\")\n",
        "\n",
        "keyword_group_raw=keyword_features\n",
        "keyword_group_raw"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>group</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>name</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>html</th>\n",
              "      <td>6</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>css</th>\n",
              "      <td>6</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>hibernate</th>\n",
              "      <td>8</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>spring</th>\n",
              "      <td>8</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>ruby</th>\n",
              "      <td>3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>perl</th>\n",
              "      <td>13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>cloud</th>\n",
              "      <td>9</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>photoshop</th>\n",
              "      <td>6</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>powershell</th>\n",
              "      <td>5</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>matlab</th>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>115 rows × 1 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "            group\n",
              "name             \n",
              "html            6\n",
              "css             6\n",
              "hibernate       8\n",
              "spring          8\n",
              "ruby            3\n",
              "...           ...\n",
              "perl           13\n",
              "cloud           9\n",
              "photoshop       6\n",
              "powershell      5\n",
              "matlab          1\n",
              "\n",
              "[115 rows x 1 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "e9k4j9bsDGas",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 475
        },
        "outputId": "b230ba13-bc98-456f-b87d-2a6f1b2ba179"
      },
      "source": [
        "#Entering node features as One-hot encoding\n",
        "keyword_features=pd.get_dummies(keyword_features,columns=[\"group\"])\n",
        "keyword_group=keyword_features\n",
        "keyword_group"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>group_1</th>\n",
              "      <th>group_2</th>\n",
              "      <th>group_3</th>\n",
              "      <th>group_4</th>\n",
              "      <th>group_5</th>\n",
              "      <th>group_6</th>\n",
              "      <th>group_7</th>\n",
              "      <th>group_8</th>\n",
              "      <th>group_9</th>\n",
              "      <th>group_10</th>\n",
              "      <th>group_11</th>\n",
              "      <th>group_12</th>\n",
              "      <th>group_13</th>\n",
              "      <th>group_14</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>name</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>html</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>css</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>hibernate</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>spring</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>ruby</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>perl</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>cloud</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>photoshop</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>powershell</th>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>matlab</th>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>115 rows × 14 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "            group_1  group_2  group_3  ...  group_12  group_13  group_14\n",
              "name                                   ...                              \n",
              "html              0        0        0  ...         0         0         0\n",
              "css               0        0        0  ...         0         0         0\n",
              "hibernate         0        0        0  ...         0         0         0\n",
              "spring            0        0        0  ...         0         0         0\n",
              "ruby              0        0        1  ...         0         0         0\n",
              "...             ...      ...      ...  ...       ...       ...       ...\n",
              "perl              0        0        0  ...         0         1         0\n",
              "cloud             0        0        0  ...         0         0         0\n",
              "photoshop         0        0        0  ...         0         0         0\n",
              "powershell        0        0        0  ...         0         0         0\n",
              "matlab            1        0        0  ...         0         0         0\n",
              "\n",
              "[115 rows x 14 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "beny47UQr49h"
      },
      "source": [
        "#### Graph creation using StellerGraph"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UFCM5AjgFhBg"
      },
      "source": [
        "from stellargraph import StellarGraph"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PliNF_BpDl5Y",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9b1bd9f4-f1e0-47d1-de33-e087459fd668"
      },
      "source": [
        "keyword_features= StellarGraph({\"source\": keyword_features}, {\"target\": keyword_links})\n",
        "print(keyword_features.info())"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "StellarGraph: Undirected multigraph\n",
            " Nodes: 115, Edges: 490\n",
            "\n",
            " Node types:\n",
            "  source: [115]\n",
            "    Features: float32 vector, length 14\n",
            "    Edge types: source-target->source\n",
            "\n",
            " Edge types:\n",
            "    source-target->source: [490]\n",
            "        Weights: all 1 (default)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7SqF9x2br-si"
      },
      "source": [
        "#### Embedding generation using GraphSAGE"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oVlRTUsUHMQX"
      },
      "source": [
        "#Unsupervised GraphSAGE implementation starts here\n",
        "\n",
        "nodes = list(keyword_features.nodes())\n",
        "number_of_walks = 1\n",
        "length = 3\n",
        "\n",
        "unsupervised_samples = UnsupervisedSampler(\n",
        "    keyword_features, nodes=nodes, length=length, number_of_walks=number_of_walks\n",
        ")\n",
        "batch_size = 20\n",
        "epochs = 4\n",
        "num_samples = [5, 3]\n",
        "\n",
        "generator = GraphSAGELinkGenerator(keyword_features, batch_size, num_samples)\n",
        "train_gen = generator.flow(unsupervised_samples)\n",
        "\n",
        "layer_sizes = [15, 15]\n",
        "graphsage = GraphSAGE(\n",
        "    layer_sizes=layer_sizes, generator=generator, bias=True, dropout=0.0, normalize=\"l2\"\n",
        ")"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Yc5hN_txHj5a"
      },
      "source": [
        "# Build the model and expose input and output sockets of graphsage, for node pair inputs:\n",
        "x_inp, x_out = graphsage.in_out_tensors()"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Gz1B1LnmHmxe",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c466e498-c226-4f18-96b6-80f4ef7db7ff"
      },
      "source": [
        "prediction = link_classification(\n",
        "    output_dim=1, output_act=\"sigmoid\", edge_embedding_method=\"ip\"\n",
        ")(x_out)"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "link_classification: using 'ip' method to combine node embeddings into edge embeddings\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uRCfKNa9HuQE",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "6c9ccf62-c8c1-45a3-e80f-13048f0093f0"
      },
      "source": [
        "model = keras.Model(inputs=x_inp, outputs=prediction)\n",
        "\n",
        "model.compile(\n",
        "    optimizer=keras.optimizers.Adam(lr=1e-3),\n",
        "    loss=keras.losses.binary_crossentropy,\n",
        "    metrics=[keras.metrics.binary_accuracy],\n",
        ")"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/optimizer_v2/optimizer_v2.py:375: UserWarning: The `lr` argument is deprecated, use `learning_rate` instead.\n",
            "  \"The `lr` argument is deprecated, use `learning_rate` instead.\")\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6-yZjHc9HxU0",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "03897c5f-e0b8-4e34-d838-4bbf79c875fb"
      },
      "source": [
        "#Model Training\n",
        "history = model.fit(\n",
        "    train_gen,\n",
        "    epochs=epochs,\n",
        "    verbose=1,\n",
        "    use_multiprocessing=False,\n",
        "    workers=4,\n",
        "    shuffle=True,\n",
        ")"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/4\n",
            "23/23 [==============================] - 2s 8ms/step - loss: 0.6774 - binary_accuracy: 0.5087\n",
            "Epoch 2/4\n",
            "23/23 [==============================] - 0s 9ms/step - loss: 0.5957 - binary_accuracy: 0.6152\n",
            "Epoch 3/4\n",
            "23/23 [==============================] - 0s 11ms/step - loss: 0.5807 - binary_accuracy: 0.7022\n",
            "Epoch 4/4\n",
            "23/23 [==============================] - 0s 10ms/step - loss: 0.5519 - binary_accuracy: 0.7391\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jGOmNY91H2o4"
      },
      "source": [
        "from sklearn.decomposition import PCA\n",
        "from sklearn.manifold import TSNE\n",
        "from stellargraph.mapper import GraphSAGENodeGenerator\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "%matplotlib inline"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lXSJh270IJLN"
      },
      "source": [
        "x_inp_src = x_inp[0::2]\n",
        "x_out_src = x_out[0]\n",
        "embedding_model = keras.Model(inputs=x_inp_src, outputs=x_out_src)"
      ],
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jPO9gTFcIMTp"
      },
      "source": [
        "node_ids = keyword_group.index\n",
        "node_gen = GraphSAGENodeGenerator(keyword_features, batch_size, num_samples).flow(node_ids)"
      ],
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JF01ibzmJdM_",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "a900cc76-eabf-4eb6-bd9a-a292939d757b"
      },
      "source": [
        "node_embeddings = embedding_model.predict(node_gen, workers=4, verbose=1)"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "6/6 [==============================] - 0s 15ms/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WVW1KPseJj0X",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "46a91ec5-71eb-46b2-ba3e-435af9c48567"
      },
      "source": [
        "node_embeddings.shape"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(115, 15)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "91rVO854rkyu"
      },
      "source": [
        "#### Clustering using embedding"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5MIa3uXqqf2H"
      },
      "source": [
        "!pip install hdbscan\n",
        "import hdbscan"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qPPSyTwGqb6w",
        "outputId": "a1601d8c-a811-48e6-f675-d81be05c6b7b"
      },
      "source": [
        "cluster = hdbscan.HDBSCAN(min_cluster_size=5, min_samples=1, metric='euclidean', cluster_selection_method='eom').fit(node_embeddings)\n",
        "res = list(zip(list(keyword_group.index),cluster.labels_))\n",
        "cluster.labels_"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([ 3,  3,  5,  5,  1,  1,  6,  6,  3,  2,  2,  4,  4,  6,  3,  3,  1,\n",
              "        1,  3,  3,  5,  4,  1,  5,  3,  3, -1, -1,  4,  1,  1,  1,  6,  7,\n",
              "        7, -1, -1,  4,  4,  4,  4,  6,  5, -1,  3,  2,  2, -1,  6,  7,  7,\n",
              "       -1,  3,  3,  4,  4,  3,  3,  2,  3,  5,  5,  5,  7,  7,  2,  2,  3,\n",
              "       -1, -1,  3,  5,  5, -1, -1,  6,  1, -1,  0,  0,  2,  7,  2,  2,  7,\n",
              "        7,  3,  1,  3,  2,  2,  0,  4,  1,  4,  7,  5,  4,  4,  0,  3,  7,\n",
              "        3,  4,  4,  4, -1,  5,  3,  4,  4,  0,  3,  7,  2])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 32
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kFl026fgrVFJ",
        "outputId": "fbd515d7-1f4f-440c-ea01-03c27c310103"
      },
      "source": [
        "cluster_dict = {}\n",
        "for i in range(-1,8):\n",
        "  l = []\n",
        "  for each in res:\n",
        "    if each[1]==i:\n",
        "      l.append(each[0])\n",
        "  cluster_dict[i] = l\n",
        "cluster_dict"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{-1: ['hadoop',\n",
              "  'apache-spark',\n",
              "  'excel',\n",
              "  'excel-vba',\n",
              "  'scala',\n",
              "  'vba',\n",
              "  'angularjs',\n",
              "  'tdd',\n",
              "  'agile',\n",
              "  'testing',\n",
              "  'selenium',\n",
              "  'jenkins',\n",
              "  'haskell'],\n",
              " 0: ['docker', 'amazon-web-services', 'go', 'devops', 'cloud'],\n",
              " 1: ['ruby',\n",
              "  'ruby-on-rails',\n",
              "  'redux',\n",
              "  'reactjs',\n",
              "  'react-native',\n",
              "  'express',\n",
              "  'node.js',\n",
              "  'mongodb',\n",
              "  'redis',\n",
              "  'elasticsearch',\n",
              "  'postgresql'],\n",
              " 2: ['c',\n",
              "  'c++',\n",
              "  'django',\n",
              "  'python',\n",
              "  'flask',\n",
              "  'angular2',\n",
              "  'typescript',\n",
              "  'angular',\n",
              "  'machine-learning',\n",
              "  'qt',\n",
              "  'r',\n",
              "  'embedded',\n",
              "  'matlab'],\n",
              " 3: ['html',\n",
              "  'css',\n",
              "  'html5',\n",
              "  'javascript',\n",
              "  'jquery',\n",
              "  'php',\n",
              "  'mysql',\n",
              "  'less',\n",
              "  'sass',\n",
              "  'ajax',\n",
              "  'asp.net-web-api',\n",
              "  'laravel',\n",
              "  'json',\n",
              "  'xml',\n",
              "  'wordpress',\n",
              "  'codeigniter',\n",
              "  'twitter-bootstrap',\n",
              "  'ionic-framework',\n",
              "  'vue.js',\n",
              "  'drupal',\n",
              "  'bootstrap',\n",
              "  'twitter-bootstrap-3',\n",
              "  'photoshop'],\n",
              " 4: ['asp.net',\n",
              "  'c#',\n",
              "  '.net',\n",
              "  'sql-server',\n",
              "  'entity-framework',\n",
              "  'linq',\n",
              "  'wcf',\n",
              "  'wpf',\n",
              "  'plsql',\n",
              "  'oracle',\n",
              "  'visual-studio',\n",
              "  'sql',\n",
              "  'vb.net',\n",
              "  'unity3d',\n",
              "  'xamarin',\n",
              "  'azure',\n",
              "  'mvc',\n",
              "  'regex',\n",
              "  'perl'],\n",
              " 5: ['hibernate',\n",
              "  'spring',\n",
              "  'spring-mvc',\n",
              "  'spring-boot',\n",
              "  'java',\n",
              "  'java-ee',\n",
              "  'maven',\n",
              "  'jsp',\n",
              "  'web-services',\n",
              "  'rest',\n",
              "  'eclipse',\n",
              "  'api'],\n",
              " 6: ['ios',\n",
              "  'swift',\n",
              "  'objective-c',\n",
              "  'iphone',\n",
              "  'android',\n",
              "  'xcode',\n",
              "  'android-studio'],\n",
              " 7: ['github',\n",
              "  'git',\n",
              "  'apache',\n",
              "  'nginx',\n",
              "  'bash',\n",
              "  'linux',\n",
              "  'osx',\n",
              "  'windows',\n",
              "  'ubuntu',\n",
              "  'unix',\n",
              "  'shell',\n",
              "  'powershell']}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4YJFj4YFsGCC"
      },
      "source": [
        "#### Visualization [Dimension reduction using PCA]"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sgQl0h-lM1Hk"
      },
      "source": [
        "X = node_embeddings\n",
        "if X.shape[1] > 2:\n",
        "    transform = TSNE  # PCA\n",
        "\n",
        "    trans = transform(n_components=2)\n",
        "    emb_transformed = pd.DataFrame(trans.fit_transform(X), index=node_ids)\n",
        "    emb_transformed[\"label\"] = keyword_group_raw\n",
        "else:\n",
        "    emb_transformed = pd.DataFrame(X, index=node_ids)\n",
        "    emb_transformed = emb_transformed.rename(columns={\"0\": 0, \"1\": 1})\n",
        "    emb_transformed[\"label\"] = keyword_group_raw"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9afil7TTTGIj",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 462
        },
        "outputId": "ac661fd8-0f21-4411-f7d5-9f06a22f63dd"
      },
      "source": [
        "alpha = 0.7\n",
        "\n",
        "fig, ax = plt.subplots(figsize=(7, 7))\n",
        "ax.scatter(\n",
        "    emb_transformed[0],\n",
        "    emb_transformed[1],\n",
        "    c=emb_transformed[\"label\"].astype(\"category\"),\n",
        "    cmap=\"jet\",\n",
        "    alpha=alpha,\n",
        ")\n",
        "ax.set(aspect=\"equal\", xlabel=\"$X_1$\", ylabel=\"$X_2$\")\n",
        "plt.title(\n",
        "    \"{} visualization of GraphSAGE embeddings for Stack Overflow Tag Network dataset\".format(transform.__name__)\n",
        ")\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 504x504 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}