12SpamDetectionWithRNN.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "12SpamDetectionWithRNN.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fITsSAPHzkTn",
        "colab_type": "text"
      },
      "source": [
        "# 1) Data Preprocessing"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WW5tpzyLvGpi",
        "colab_type": "code",
        "outputId": "5ba8051d-7bac-496b-b6b7-8359f3726fb0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 33
        }
      },
      "source": [
        "import tensorflow as tf\n",
        "print(tf.__version__)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2.2.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_bsRRcehyj6r",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "from tensorflow.keras.preprocessing.text import Tokenizer\n",
        "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
        "from tensorflow.keras.layers import Dense, Input, GlobalMaxPooling1D\n",
        "from tensorflow.keras.layers import LSTM, Embedding\n",
        "from tensorflow.keras.models import Model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IpmMS1NqyxOi",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df = pd.read_csv('spamrnn.csv', encoding='ISO-8859-1')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SNMCRunDy7jJ",
        "colab_type": "code",
        "outputId": "642a940a-1fcf-4d20-c0f9-31a3a1483b98",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 196
        }
      },
      "source": [
        "df.head()"
      ],
      "execution_count": 6,
      "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>v1</th>\n",
              "      <th>v2</th>\n",
              "      <th>Unnamed: 2</th>\n",
              "      <th>Unnamed: 3</th>\n",
              "      <th>Unnamed: 4</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>ham</td>\n",
              "      <td>Go until jurong point, crazy.. Available only ...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>ham</td>\n",
              "      <td>Ok lar... Joking wif u oni...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>spam</td>\n",
              "      <td>Free entry in 2 a wkly comp to win FA Cup fina...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>ham</td>\n",
              "      <td>U dun say so early hor... U c already then say...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>ham</td>\n",
              "      <td>Nah I don't think he goes to usf, he lives aro...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "     v1  ... Unnamed: 4\n",
              "0   ham  ...        NaN\n",
              "1   ham  ...        NaN\n",
              "2  spam  ...        NaN\n",
              "3   ham  ...        NaN\n",
              "4   ham  ...        NaN\n",
              "\n",
              "[5 rows x 5 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3mX-Z5RRzYrl",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# drop garbage columns\n",
        "df = df.drop([\"Unnamed: 2\", \"Unnamed: 3\", \"Unnamed: 4\"], axis=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AcazYbp8zcG1",
        "colab_type": "code",
        "outputId": "e3ba7a0e-ef37-490f-dfd3-4482cec04220",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 196
        }
      },
      "source": [
        "df.head()"
      ],
      "execution_count": 8,
      "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>v1</th>\n",
              "      <th>v2</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>ham</td>\n",
              "      <td>Go until jurong point, crazy.. Available only ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>ham</td>\n",
              "      <td>Ok lar... Joking wif u oni...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>spam</td>\n",
              "      <td>Free entry in 2 a wkly comp to win FA Cup fina...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>ham</td>\n",
              "      <td>U dun say so early hor... U c already then say...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>ham</td>\n",
              "      <td>Nah I don't think he goes to usf, he lives aro...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "     v1                                                 v2\n",
              "0   ham  Go until jurong point, crazy.. Available only ...\n",
              "1   ham                      Ok lar... Joking wif u oni...\n",
              "2  spam  Free entry in 2 a wkly comp to win FA Cup fina...\n",
              "3   ham  U dun say so early hor... U c already then say...\n",
              "4   ham  Nah I don't think he goes to usf, he lives aro..."
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hcWkSMuIzhRd",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# rename columns\n",
        "df.columns = ['labels', 'data']"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mKuCO0rfzigW",
        "colab_type": "code",
        "outputId": "1e2ecec9-9463-4f5d-b361-3748d1179dd5",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 196
        }
      },
      "source": [
        "df.head()"
      ],
      "execution_count": 10,
      "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>labels</th>\n",
              "      <th>data</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>ham</td>\n",
              "      <td>Go until jurong point, crazy.. Available only ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>ham</td>\n",
              "      <td>Ok lar... Joking wif u oni...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>spam</td>\n",
              "      <td>Free entry in 2 a wkly comp to win FA Cup fina...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>ham</td>\n",
              "      <td>U dun say so early hor... U c already then say...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>ham</td>\n",
              "      <td>Nah I don't think he goes to usf, he lives aro...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  labels                                               data\n",
              "0    ham  Go until jurong point, crazy.. Available only ...\n",
              "1    ham                      Ok lar... Joking wif u oni...\n",
              "2   spam  Free entry in 2 a wkly comp to win FA Cup fina...\n",
              "3    ham  U dun say so early hor... U c already then say...\n",
              "4    ham  Nah I don't think he goes to usf, he lives aro..."
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PaQ_aA-qzsDV",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# create binary labels (0 and 1)\n",
        "df['b_labels'] = df['labels'].map({'ham': 0, 'spam': 1}) # create new column as 'b_labels'\n",
        "y = df['b_labels'].values"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vguK3gr2zsjC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# split the data\n",
        "x_train, x_test, y_train, y_test = train_test_split(df['data'], y, test_size=0.33)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uoYIycZPasKv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Convert sentences to sequences\n",
        "max_vocab_size = 20000\n",
        "tokenizer = Tokenizer(num_words=max_vocab_size)\n",
        "tokenizer.fit_on_texts(x_train)\n",
        "sequences_train = tokenizer.texts_to_sequences(x_train)\n",
        "sequences_test = tokenizer.texts_to_sequences(x_test)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5U7Mv9nXatkz",
        "colab_type": "code",
        "outputId": "12b5192b-bc11-4935-f511-4b42ceb7670e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 33
        }
      },
      "source": [
        "# Check word index mapping (to check the number of words in vocabulary)\n",
        "word2idx = tokenizer.word_index\n",
        "V = len(word2idx)\n",
        "print('Total number of unique tokens are: %s' % V)"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Total number of unique tokens are: 7114\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OMJZ8q4TgBK4",
        "colab_type": "code",
        "outputId": "f6a2cdc3-f57c-4cf8-c99c-e0f9d2b4baa1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 33
        }
      },
      "source": [
        "# pad sequences (to get  N x T matrix)\n",
        "data_train = pad_sequences(sequences_train)\n",
        "print('Shape of data train tensor:', data_train.shape)\n",
        "\n",
        "# N: number of samples and T: Number of time steps"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Shape of data train tensor: (3733, 189)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Q79uMI-ENWiP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# set the value of T to get sequence length\n",
        "T = data_train.shape[1]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_4unHL2hg-vg",
        "colab_type": "code",
        "outputId": "be01167c-cccb-48e2-ad15-f30ed6fb552d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 33
        }
      },
      "source": [
        "print(T)"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "189\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OCerX6JqgFQ9",
        "colab_type": "code",
        "outputId": "aa2141e3-d561-4e85-ee13-1bef3d3a04bc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 33
        }
      },
      "source": [
        "# pad the test set\n",
        "data_test = pad_sequences(sequences_test, maxlen=T)\n",
        "# maxlen = T, to truncate longer sentences in test set\n",
        "print('Shape of data test tensor:', data_test.shape)"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Shape of data test tensor: (1839, 189)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_HHnMf7lhjFI",
        "colab_type": "text"
      },
      "source": [
        "# 2) Building the Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nEXwk99KgJri",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "\n",
        "# Choose embedding dimensionality\n",
        "D = 20  # this is a hyper parameter, we can choose any word vector size that we want\n",
        "\n",
        "# Hidden state vectorsize (dimensionality)\n",
        "M = 15\n",
        "\n",
        "# Input layer\n",
        "i = Input(shape=(T,))  # input layer takes in sequences of integers, so shape is T\n",
        "\n",
        "# Embedding layer\n",
        "x = Embedding(V + 1, D)(i) # This takes in sequences of integers and returns sequences of word vectors\n",
        "# This will be an N * T * D array\n",
        "# we want size of embedding to (V + 1) x D, because first word index starts from 1 and not 0\n",
        "\n",
        "# LSTM layer\n",
        "x = LSTM(M, return_sequences=True)(x)\n",
        "x = GlobalMaxPooling1D()(x)\n",
        "\n",
        "# Dense layer\n",
        "x = Dense(1, activation='sigmoid')(x)\n",
        "# it is an binary classification problem, so we are using activation function ='sigmoid'\n",
        "\n",
        "model = Model(i, x)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a9D9_efOgKEI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# Compile the model\n",
        "model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Mettjf8smRse",
        "colab_type": "code",
        "outputId": "1ef99359-2541-4798-9196-2d2ef145de5b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 371
        }
      },
      "source": [
        "# Train the model\n",
        "r = model.fit(x=data_train, y=y_train, epochs=10, validation_data=(data_test, y_test))"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "117/117 [==============================] - 2s 17ms/step - loss: 0.4565 - accuracy: 0.8685 - val_loss: 0.4047 - val_accuracy: 0.8608\n",
            "Epoch 2/10\n",
            "117/117 [==============================] - 2s 13ms/step - loss: 0.3687 - accuracy: 0.8685 - val_loss: 0.3003 - val_accuracy: 0.8608\n",
            "Epoch 3/10\n",
            "117/117 [==============================] - 2s 13ms/step - loss: 0.2337 - accuracy: 0.8685 - val_loss: 0.2096 - val_accuracy: 0.8608\n",
            "Epoch 4/10\n",
            "117/117 [==============================] - 2s 13ms/step - loss: 0.1636 - accuracy: 0.9392 - val_loss: 0.1634 - val_accuracy: 0.9717\n",
            "Epoch 5/10\n",
            "117/117 [==============================] - 2s 13ms/step - loss: 0.1220 - accuracy: 0.9882 - val_loss: 0.1310 - val_accuracy: 0.9766\n",
            "Epoch 6/10\n",
            "117/117 [==============================] - 2s 14ms/step - loss: 0.0916 - accuracy: 0.9922 - val_loss: 0.1078 - val_accuracy: 0.9853\n",
            "Epoch 7/10\n",
            "117/117 [==============================] - 2s 13ms/step - loss: 0.0682 - accuracy: 0.9960 - val_loss: 0.0897 - val_accuracy: 0.9859\n",
            "Epoch 8/10\n",
            "117/117 [==============================] - 2s 14ms/step - loss: 0.0494 - accuracy: 0.9965 - val_loss: 0.0796 - val_accuracy: 0.9859\n",
            "Epoch 9/10\n",
            "117/117 [==============================] - 2s 13ms/step - loss: 0.0380 - accuracy: 0.9971 - val_loss: 0.0739 - val_accuracy: 0.9837\n",
            "Epoch 10/10\n",
            "117/117 [==============================] - 2s 13ms/step - loss: 0.0301 - accuracy: 0.9979 - val_loss: 0.0712 - val_accuracy: 0.9826\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lk8FQntigMXW",
        "colab_type": "code",
        "outputId": "4df1dc70-f050-4d63-abe3-ca0c4771e91e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        }
      },
      "source": [
        "# Loss per iteration\n",
        "import matplotlib.pyplot as plt\n",
        "plt.plot(r.history['loss'], label='Loss')\n",
        "plt.plot(r.history['val_loss'], label='Validation Loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gYYsAuRJgOJn",
        "colab_type": "code",
        "outputId": "e18599d3-e150-4e9b-ccaf-d95834860db2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        }
      },
      "source": [
        "# Accuracy per iteration\n",
        "plt.plot(r.history['accuracy'], label='Accuracy')\n",
        "plt.plot(r.history['val_accuracy'], label='Validation accuracy')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}