rnn.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Copy of rnn.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xpAuMVCwfWs8"
      },
      "source": [
        "# Recurrent Neural Network"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ir9zwETrfbrp"
      },
      "source": [
        "### Importing the libraries"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZT1f24vHffuf"
      },
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nQ47JAxrgmaL"
      },
      "source": [
        "### Importing the training set"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1xiv3pJOgqY3"
      },
      "source": [
        "dataset_train = pd.read_csv('/content/Uniqlo(FastRetailing) 2012-2016 Training - stocks2012-2016.csv')\n",
        "training_set = dataset_train.iloc[:, 1:2].values"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HT8_2UJegtG5"
      },
      "source": [
        "### Feature Scaling"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OTrF2kR7gx9x"
      },
      "source": [
        "from sklearn.preprocessing import MinMaxScaler\n",
        "sc = MinMaxScaler(feature_range = (0, 1))\n",
        "training_set_scaled = sc.fit_transform(training_set)"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JyYgYocqhNUg"
      },
      "source": [
        "### Creating a data structure with 60 timesteps and 1 output"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iofU21B0i6ST"
      },
      "source": [
        "X_train = []\n",
        "y_train = []\n",
        "for i in range(60, 1226):\n",
        "    X_train.append(training_set_scaled[i-60:i, 0])\n",
        "    y_train.append(training_set_scaled[i, 0])\n",
        "X_train, y_train = np.array(X_train), np.array(y_train)"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "D8yaN7Zvi95l"
      },
      "source": [
        "### Reshaping"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FOXqJHmNjBkz"
      },
      "source": [
        "X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1))"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YfavHaDL2biz",
        "outputId": "06d51495-3bc0-45a0-d1cb-9e1a3f50aa6a"
      },
      "source": [
        "X_train.dtype"
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "dtype('float64')"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k4XV88JMjHXG"
      },
      "source": [
        "### Importing the Keras libraries and packages"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9JRnqsxEjKsD"
      },
      "source": [
        "from keras.models import Sequential\n",
        "from keras.layers import Dense\n",
        "from keras.layers import LSTM\n",
        "from keras.layers import Dropout"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FEIE-1s9jNzC"
      },
      "source": [
        "### Initialising the RNN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1338dJ0UjRKH"
      },
      "source": [
        "rnn_model = Sequential()"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "62eg1OPGjT8z"
      },
      "source": [
        "### Adding the first LSTM layer and some Dropout regularisation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2hIinyXUjbVU"
      },
      "source": [
        "rnn_model.add(LSTM(units = 50, return_sequences = True, input_shape = (X_train.shape[1], 1)))\n",
        "rnn_model.add(Dropout(0.2))"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3XBIYLyOjlMx"
      },
      "source": [
        "### Adding a second LSTM layer and some Dropout regularisation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UG7nrVaSjuZ2"
      },
      "source": [
        "rnn_model.add(LSTM(units = 50, return_sequences = True))\n",
        "rnn_model.add(Dropout(0.2))"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ey3fHVnGj1cu"
      },
      "source": [
        "### Adding a third LSTM layer and some Dropout regularisation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PuNi6PgFj7jO"
      },
      "source": [
        "rnn_model.add(LSTM(units = 50, return_sequences = True))\n",
        "rnn_model.add(Dropout(0.2))"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SYTrtfTmj933"
      },
      "source": [
        "### Adding a fourth LSTM layer and some Dropout regularisation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Jp4Ty8fRkBYV"
      },
      "source": [
        "rnn_model.add(LSTM(units = 50))\n",
        "rnn_model.add(Dropout(0.2))"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9ABI6rOIkHhk"
      },
      "source": [
        "### Adding the output layer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aM6R1z4WkME8"
      },
      "source": [
        "rnn_model.add(Dense(units = 1))"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zLx4K7uUkPSh"
      },
      "source": [
        "### Compiling the RNN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XTrhVN-tkbko"
      },
      "source": [
        "rnn_model.compile(optimizer = 'adam', loss = 'mean_squared_error')"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-mPhwKGkkebi"
      },
      "source": [
        "### Fitting the RNN to the Training set"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I06Nkrz5kkb-",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7345a541-763e-4171-943c-2aae8d4100c9"
      },
      "source": [
        "rnn_model.fit(X_train, y_train, epochs = 100, batch_size = 32)"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "37/37 [==============================] - 11s 112ms/step - loss: 0.0292\n",
            "Epoch 2/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0058\n",
            "Epoch 3/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0055\n",
            "Epoch 4/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0056\n",
            "Epoch 5/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0044\n",
            "Epoch 6/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0051\n",
            "Epoch 7/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0046\n",
            "Epoch 8/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0048\n",
            "Epoch 9/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0045\n",
            "Epoch 10/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0044\n",
            "Epoch 11/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0039\n",
            "Epoch 12/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0039\n",
            "Epoch 13/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0039\n",
            "Epoch 14/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0038\n",
            "Epoch 15/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0035\n",
            "Epoch 16/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0035\n",
            "Epoch 17/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0037\n",
            "Epoch 18/100\n",
            "37/37 [==============================] - 4s 110ms/step - loss: 0.0035\n",
            "Epoch 19/100\n",
            "37/37 [==============================] - 4s 110ms/step - loss: 0.0030\n",
            "Epoch 20/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0030\n",
            "Epoch 21/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0030\n",
            "Epoch 22/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0027\n",
            "Epoch 23/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0030\n",
            "Epoch 24/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0029\n",
            "Epoch 25/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0028\n",
            "Epoch 26/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0025\n",
            "Epoch 27/100\n",
            "37/37 [==============================] - 4s 110ms/step - loss: 0.0025\n",
            "Epoch 28/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0026\n",
            "Epoch 29/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0028\n",
            "Epoch 30/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0026\n",
            "Epoch 31/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0026\n",
            "Epoch 32/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0026\n",
            "Epoch 33/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0026\n",
            "Epoch 34/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0023\n",
            "Epoch 35/100\n",
            "37/37 [==============================] - 4s 110ms/step - loss: 0.0023\n",
            "Epoch 36/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0024\n",
            "Epoch 37/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0021\n",
            "Epoch 38/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0021\n",
            "Epoch 39/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0022\n",
            "Epoch 40/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0021\n",
            "Epoch 41/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0020\n",
            "Epoch 42/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0020\n",
            "Epoch 43/100\n",
            "37/37 [==============================] - 4s 114ms/step - loss: 0.0020\n",
            "Epoch 44/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0018\n",
            "Epoch 45/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0022\n",
            "Epoch 46/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0021\n",
            "Epoch 47/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0019\n",
            "Epoch 48/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0016\n",
            "Epoch 49/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0016\n",
            "Epoch 50/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0018\n",
            "Epoch 51/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0019\n",
            "Epoch 52/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0016\n",
            "Epoch 53/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0018\n",
            "Epoch 54/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0017\n",
            "Epoch 55/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0015\n",
            "Epoch 56/100\n",
            "37/37 [==============================] - 4s 110ms/step - loss: 0.0016\n",
            "Epoch 57/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0017\n",
            "Epoch 58/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0015\n",
            "Epoch 59/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0016\n",
            "Epoch 60/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0016\n",
            "Epoch 61/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0017\n",
            "Epoch 62/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0016\n",
            "Epoch 63/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0014\n",
            "Epoch 64/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0016\n",
            "Epoch 65/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0016\n",
            "Epoch 66/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0016\n",
            "Epoch 67/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0013\n",
            "Epoch 68/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0014\n",
            "Epoch 69/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0014\n",
            "Epoch 70/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0013\n",
            "Epoch 71/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0014\n",
            "Epoch 72/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0013\n",
            "Epoch 73/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0014\n",
            "Epoch 74/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0013\n",
            "Epoch 75/100\n",
            "37/37 [==============================] - 4s 114ms/step - loss: 0.0011\n",
            "Epoch 76/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0012\n",
            "Epoch 77/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0013\n",
            "Epoch 78/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0013\n",
            "Epoch 79/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0013\n",
            "Epoch 80/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0014\n",
            "Epoch 81/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0014\n",
            "Epoch 82/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0014\n",
            "Epoch 83/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0013\n",
            "Epoch 84/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0012\n",
            "Epoch 85/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0012\n",
            "Epoch 86/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0015\n",
            "Epoch 87/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0013\n",
            "Epoch 88/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0013\n",
            "Epoch 89/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0011\n",
            "Epoch 90/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0012\n",
            "Epoch 91/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0011\n",
            "Epoch 92/100\n",
            "37/37 [==============================] - 4s 111ms/step - loss: 0.0011\n",
            "Epoch 93/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0011\n",
            "Epoch 94/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0011\n",
            "Epoch 95/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0010\n",
            "Epoch 96/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0012\n",
            "Epoch 97/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0011\n",
            "Epoch 98/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0011\n",
            "Epoch 99/100\n",
            "37/37 [==============================] - 4s 112ms/step - loss: 0.0011\n",
            "Epoch 100/100\n",
            "37/37 [==============================] - 4s 113ms/step - loss: 0.0012\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f0ac5183390>"
            ]
          },
          "metadata": {},
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "SgJO6qEDksxD"
      },
      "source": [
        "### Getting the real stock price of 2017"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FmBT2zqukxTz"
      },
      "source": [
        "dataset_test = pd.read_csv('/content/Uniqlo(FastRetailing) 2017 Test - stocks2017.csv')\n",
        "real_stock_price = dataset_test.iloc[:, 1:2].values"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GrvrLblxkz42"
      },
      "source": [
        "### Getting the predicted stock price of 2017"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "emikTvUpk3Ck",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ff2dec40-7165-464e-b0d6-56b5d74f21a7"
      },
      "source": [
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\", category=np.VisibleDeprecationWarning) \n",
        "dataset_total = pd.concat((dataset_train['Open'], dataset_test['Open']), axis = 0)\n",
        "inputs = dataset_total[len(dataset_total) - len(dataset_test) - 60:].values\n",
        "print(inputs.shape)\n",
        "inputs = inputs.reshape(-1,1)\n",
        "inputs = sc.transform(inputs)\n",
        "X_test = []\n",
        "for i in range(60, 68):\n",
        "    X_test.append(inputs[i-60:i, 0])\n",
        "    \n",
        "#print(len(X_test), len(X_test[8]))\n",
        "X_test = np.array(X_test)\n",
        "#X_test = np.array(X_test, dtype='float64')\n",
        "\n",
        "print(X_test.shape)\n",
        "X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1))\n",
        "predicted_stock_price = regressor.predict(X_test)\n",
        "predicted_stock_price = sc.inverse_transform(predicted_stock_price)"
      ],
      "execution_count": 58,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(67,)\n",
            "(8, 60)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iFTNs3YHk6FQ"
      },
      "source": [
        "### Visualising the results"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8OUI8U49k9tH",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "a55d22c0-e97c-4c26-8a7c-c2cac396ed72"
      },
      "source": [
        "plt.plot(real_stock_price, color = 'red', label = 'Real Uniqlo Stock Price')\n",
        "plt.plot(predicted_stock_price, color = 'blue', label = 'Predicted Uniqlo Stock Price')\n",
        "plt.title('Uniqlo Stock Price Prediction')\n",
        "plt.xlabel('Time')\n",
        "plt.ylabel('Uniqlo Stock Price')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 59,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}