Context_Neural_NN_w_early_stopping.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Context_Neural_NN_w_early_stopping.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "A5U7t4YqnxgG",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#!/usr/bin/env python\n",
        "# coding: utf-8\n",
        "\n",
        "# In[4]:\n",
        "\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "from sklearn.metrics import r2_score\n",
        "\n",
        "\n",
        "# In[5]:\n",
        "\n",
        "\n",
        "import keras\n",
        "\n",
        "\n",
        "# In[61]:\n",
        "\n",
        "\n",
        "# Keras\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Flatten, LSTM, Dropout, Activation\n",
        "from keras import backend as K\n",
        "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
        "\n",
        "np.random.seed(1)     \n",
        "\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "o6q9OecKoAVq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#TEST_NAME = \"KMeans\"\n",
        "TEST_NAME = \"PageRank\"\n",
        "#TEST_NAME = \"SGD\"\n",
        "\n",
        "#plt.style.use('ggplot')\n",
        "\n",
        "def plot_history(history):\n",
        "    #acc = history.history['r2_score_other']\n",
        "    #val_acc = history.history['val_acc']\n",
        "    loss = history.history['loss']\n",
        "    val_loss = history.history['val_loss']\n",
        "    x = range(1, len(loss) + 1)\n",
        "\n",
        "    #plt.figure(figsize=(12, 5))\n",
        "    #plt.subplot(1, 2, 1)\n",
        "    #plt.plot(x, acc, 'b', label='Training r2_score')\n",
        "    #plt.plot(x, val_acc, 'r', label='Validation acc')\n",
        "    #plt.title('Training and validation accuracy')\n",
        "    #plt.legend()\n",
        "    #plt.subplot(1, 2, 2)\n",
        "    plt.plot(x, loss, 'b', label='Training loss')\n",
        "    plt.plot(x, val_loss, 'r', label='Validation loss')\n",
        "    plt.title('Training and validation loss')\n",
        "    plt.legend()\n",
        "\n",
        "\n",
        "# In[42]:\n",
        "\n",
        "def add_context(data_df, context = 0):\n",
        "\n",
        "    L = len(data_df)\n",
        "    context_data = []\n",
        "    row_1 = pd.DataFrame(data_df.iloc[0])\n",
        "    #pad data with data[0] context times\n",
        "    data = pd.concat([row_1]*context + [data_df])\n",
        "    data = data_df.to_numpy()\n",
        "    cols = list(data_df.columns)\n",
        "\n",
        "    for fi in range(context, L):\n",
        "        frame = data[fi-context:fi+1].flatten()\n",
        "        context_data.append(frame)\n",
        "    return pd.DataFrame(context_data, columns = cols*(context+1))\n",
        "\n",
        "def read_dataset(filename):\n",
        "    dataset = pd.read_csv(filename)\n",
        "    df_cols = list(dataset.columns)\n",
        "    dataset = scaler.fit_transform(dataset)\n",
        "    dataset = pd.DataFrame(dataset, columns = df_cols)\n",
        "    return dataset\n",
        "\n",
        "def get_data_labels(df, context = 0):\n",
        "    label = df['flow_size'][:-context]\n",
        "    df = add_context(df, context)\n",
        "    df.iloc[:, -1] = 0\n",
        "    #add context to labels\n",
        "    return df, label\n",
        "\n",
        "\n",
        "\n",
        "# # Neural Network\n",
        "\n",
        "\n",
        "#from keras import optimizers\n",
        "#adam = optimizers.Adam(lr=0.9)\n",
        "\n",
        "\n",
        "# In[90]:\n",
        "\n",
        "\n",
        "#r2 score used in paper. 1 is best. -1 is bad.\n",
        "\n",
        "def r2_score_other(y_true, y_pred):\n",
        "    from keras import backend as K\n",
        "    SS_res =  K.sum(K.square( y_true-y_pred ))\n",
        "    SS_tot = K.sum(K.square( y_true - K.mean(y_true) ) )\n",
        "    return ( 1 - SS_res/(SS_tot + K.epsilon()) )\n",
        "\n",
        "\n",
        "#from https://stackoverflow.com/questions/45250100/kerasregressor-coefficient-of-determination-r2-score"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r5ePTgxooHuF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "scaler = MinMaxScaler(feature_range=(0, 1))\n",
        "\n",
        "train_df = read_dataset(TEST_NAME+'_training.csv')\n",
        "test_df = read_dataset(TEST_NAME+'_test.csv')\n",
        "validation_df = read_dataset(TEST_NAME+'_validation.csv')\n",
        "\n",
        "context = 1\n",
        "\n",
        "train_x, train_y = get_data_labels(train_df, context)\n",
        "test_x, test_y = get_data_labels(test_df, context)\n",
        "validation_x, validation_y = get_data_labels(validation_df, context)\n",
        "#val = (val_x, val_y)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WcvJZ6NO097T",
        "colab_type": "code",
        "outputId": "f0375598-b85e-4dd7-d615-5d2c3fa2ea24",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 299
        }
      },
      "source": [
        "#train_x.head()"
      ],
      "execution_count": 106,
      "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>time</th>\n",
              "      <th>flow_frequency</th>\n",
              "      <th>agg_net_out</th>\n",
              "      <th>agg_net_in</th>\n",
              "      <th>agg_net_out_per_machine</th>\n",
              "      <th>agg_net_in_per_machine</th>\n",
              "      <th>machine</th>\n",
              "      <th>flow_size</th>\n",
              "      <th>time</th>\n",
              "      <th>flow_frequency</th>\n",
              "      <th>agg_net_out</th>\n",
              "      <th>agg_net_in</th>\n",
              "      <th>agg_net_out_per_machine</th>\n",
              "      <th>agg_net_in_per_machine</th>\n",
              "      <th>machine</th>\n",
              "      <th>flow_size</th>\n",
              "      <th>time</th>\n",
              "      <th>flow_frequency</th>\n",
              "      <th>agg_net_out</th>\n",
              "      <th>agg_net_in</th>\n",
              "      <th>agg_net_out_per_machine</th>\n",
              "      <th>agg_net_in_per_machine</th>\n",
              "      <th>machine</th>\n",
              "      <th>flow_size</th>\n",
              "      <th>time</th>\n",
              "      <th>flow_frequency</th>\n",
              "      <th>agg_net_out</th>\n",
              "      <th>agg_net_in</th>\n",
              "      <th>agg_net_out_per_machine</th>\n",
              "      <th>agg_net_in_per_machine</th>\n",
              "      <th>machine</th>\n",
              "      <th>flow_size</th>\n",
              "      <th>time</th>\n",
              "      <th>flow_frequency</th>\n",
              "      <th>agg_net_out</th>\n",
              "      <th>agg_net_in</th>\n",
              "      <th>agg_net_out_per_machine</th>\n",
              "      <th>agg_net_in_per_machine</th>\n",
              "      <th>machine</th>\n",
              "      <th>flow_size</th>\n",
              "      <th>time</th>\n",
              "      <th>flow_frequency</th>\n",
              "      <th>agg_net_out</th>\n",
              "      <th>agg_net_in</th>\n",
              "      <th>agg_net_out_per_machine</th>\n",
              "      <th>agg_net_in_per_machine</th>\n",
              "      <th>machine</th>\n",
              "      <th>flow_size</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0.058482</td>\n",
              "      <td>0.065067</td>\n",
              "      <td>0.000003</td>\n",
              "      <td>6.937512e-08</td>\n",
              "      <td>0.000003</td>\n",
              "      <td>6.955599e-08</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>9.449852e-07</td>\n",
              "      <td>0.065533</td>\n",
              "      <td>0.009067</td>\n",
              "      <td>0.000007</td>\n",
              "      <td>0.000002</td>\n",
              "      <td>0.000007</td>\n",
              "      <td>0.000002</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>2.836602e-06</td>\n",
              "      <td>0.066985</td>\n",
              "      <td>0.000800</td>\n",
              "      <td>0.000015</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.000015</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>6.191793e-06</td>\n",
              "      <td>0.068022</td>\n",
              "      <td>0.001333</td>\n",
              "      <td>0.000017</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.000017</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>8.099873e-07</td>\n",
              "      <td>0.069266</td>\n",
              "      <td>0.000800</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>0.000005</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>0.000005</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>0.000012</td>\n",
              "      <td>0.070510</td>\n",
              "      <td>0.000533</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.000070</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0.065533</td>\n",
              "      <td>0.009067</td>\n",
              "      <td>0.000007</td>\n",
              "      <td>2.126614e-06</td>\n",
              "      <td>0.000007</td>\n",
              "      <td>2.132159e-06</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>2.836602e-06</td>\n",
              "      <td>0.066985</td>\n",
              "      <td>0.000800</td>\n",
              "      <td>0.000015</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.000015</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>6.191793e-06</td>\n",
              "      <td>0.068022</td>\n",
              "      <td>0.001333</td>\n",
              "      <td>0.000017</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.000017</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>8.099873e-07</td>\n",
              "      <td>0.069266</td>\n",
              "      <td>0.000800</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>0.000005</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>0.000005</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>1.170037e-05</td>\n",
              "      <td>0.070510</td>\n",
              "      <td>0.000533</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.000070</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>0.000039</td>\n",
              "      <td>0.072377</td>\n",
              "      <td>0.001867</td>\n",
              "      <td>0.000075</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000076</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0.066985</td>\n",
              "      <td>0.000800</td>\n",
              "      <td>0.000015</td>\n",
              "      <td>3.587494e-06</td>\n",
              "      <td>0.000015</td>\n",
              "      <td>3.596847e-06</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>6.191793e-06</td>\n",
              "      <td>0.068022</td>\n",
              "      <td>0.001333</td>\n",
              "      <td>0.000017</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.000017</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>8.099873e-07</td>\n",
              "      <td>0.069266</td>\n",
              "      <td>0.000800</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>0.000005</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>0.000005</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>1.170037e-05</td>\n",
              "      <td>0.070510</td>\n",
              "      <td>0.000533</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.000070</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>3.875427e-05</td>\n",
              "      <td>0.072377</td>\n",
              "      <td>0.001867</td>\n",
              "      <td>0.000075</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000076</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.076109</td>\n",
              "      <td>0.097867</td>\n",
              "      <td>0.000081</td>\n",
              "      <td>0.000084</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.500000</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0.068022</td>\n",
              "      <td>0.001333</td>\n",
              "      <td>0.000017</td>\n",
              "      <td>3.736917e-06</td>\n",
              "      <td>0.000017</td>\n",
              "      <td>3.746660e-06</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>8.099873e-07</td>\n",
              "      <td>0.069266</td>\n",
              "      <td>0.000800</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>0.000005</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>0.000005</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>1.170037e-05</td>\n",
              "      <td>0.070510</td>\n",
              "      <td>0.000533</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.000070</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>3.875427e-05</td>\n",
              "      <td>0.072377</td>\n",
              "      <td>0.001867</td>\n",
              "      <td>0.000075</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000076</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>4.211275e-06</td>\n",
              "      <td>0.076109</td>\n",
              "      <td>0.097867</td>\n",
              "      <td>0.000081</td>\n",
              "      <td>0.000084</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.500000</td>\n",
              "      <td>0.000004</td>\n",
              "      <td>0.078183</td>\n",
              "      <td>0.000533</td>\n",
              "      <td>0.000101</td>\n",
              "      <td>0.001455</td>\n",
              "      <td>0.000021</td>\n",
              "      <td>0.001374</td>\n",
              "      <td>0.500000</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0.069266</td>\n",
              "      <td>0.000800</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>5.315201e-06</td>\n",
              "      <td>0.000030</td>\n",
              "      <td>5.329059e-06</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>1.170037e-05</td>\n",
              "      <td>0.070510</td>\n",
              "      <td>0.000533</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.000070</td>\n",
              "      <td>0.000066</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>3.875427e-05</td>\n",
              "      <td>0.072377</td>\n",
              "      <td>0.001867</td>\n",
              "      <td>0.000075</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.000076</td>\n",
              "      <td>0.000069</td>\n",
              "      <td>0.666667</td>\n",
              "      <td>4.211275e-06</td>\n",
              "      <td>0.076109</td>\n",
              "      <td>0.097867</td>\n",
              "      <td>0.000081</td>\n",
              "      <td>0.000084</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.500000</td>\n",
              "      <td>4.026888e-06</td>\n",
              "      <td>0.078183</td>\n",
              "      <td>0.000533</td>\n",
              "      <td>0.000101</td>\n",
              "      <td>0.001455</td>\n",
              "      <td>0.000021</td>\n",
              "      <td>0.001374</td>\n",
              "      <td>0.500000</td>\n",
              "      <td>0.000020</td>\n",
              "      <td>0.087723</td>\n",
              "      <td>0.000533</td>\n",
              "      <td>0.000208</td>\n",
              "      <td>0.023754</td>\n",
              "      <td>0.000128</td>\n",
              "      <td>0.023731</td>\n",
              "      <td>0.500000</td>\n",
              "      <td>0.0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "       time  flow_frequency  ...   machine  flow_size\n",
              "0  0.058482        0.065067  ...  0.666667        0.0\n",
              "1  0.065533        0.009067  ...  0.666667        0.0\n",
              "2  0.066985        0.000800  ...  0.500000        0.0\n",
              "3  0.068022        0.001333  ...  0.500000        0.0\n",
              "4  0.069266        0.000800  ...  0.500000        0.0\n",
              "\n",
              "[5 rows x 48 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 106
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qfuz--xvOGwM",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#define callback for early stopping\n",
        "early_stop = EarlyStopping(monitor='val_loss', mode='min', patience=50)\n",
        "\n",
        "#define callback for saving best model\n",
        "best_model = ModelCheckpoint('nn_model_PageRank.h5', monitor='val_loss',mode='min')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_-j7xHv3oMV-",
        "colab_type": "code",
        "outputId": "ca3257e7-f67f-4d48-9c8b-3eef26542716",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "# In[102]:\n",
        "\n",
        "'''\n",
        "model = Sequential()\n",
        "model.add(Dense(5, input_dim=train_x.shape[1],activation='relu'))\n",
        "model.add(Dense(5, activation='relu'))\n",
        "model.add(Dense(1, activation='sigmoid'))\n",
        "#model.add(Dense(1, activation='relu'))\n",
        "model.summary()\n",
        "model.compile(loss='mean_absolute_error', optimizer='adam', metrics = [r2_score_other])\n",
        "model.fit(train_x, train_y, epochs=250, batch_size=10)\n",
        "'''\n",
        "\n",
        "model = Sequential()\n",
        "model.add(Dense(5, input_dim=train_x.shape[1], activation='relu'))\n",
        "model.add(Dense(5, activation='relu'))\n",
        "model.add(Dense(1, activation='sigmoid'))\n",
        "model.compile(loss='mean_absolute_error', optimizer='adam')\n",
        "history = model.fit(train_x, train_y, epochs=250, \\\n",
        "                      validation_data = (validation_x, validation_y), \\\n",
        "                      batch_size=10, verbose=2, callbacks = [early_stop, best_model])\n",
        "\n",
        "#plot_history(history)\n",
        "\n",
        "#error metric\n",
        "model.evaluate(test_x, test_y)"
      ],
      "execution_count": 104,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 54645 samples, validate on 15694 samples\n",
            "Epoch 1/250\n",
            " - 6s - loss: 0.0100 - val_loss: 0.0023\n",
            "Epoch 2/250\n",
            " - 6s - loss: 0.0020 - val_loss: 0.0023\n",
            "Epoch 3/250\n",
            " - 6s - loss: 0.0020 - val_loss: 0.0023\n",
            "Epoch 4/250\n",
            " - 6s - loss: 0.0018 - val_loss: 0.0014\n",
            "Epoch 5/250\n",
            " - 6s - loss: 0.0011 - val_loss: 0.0013\n",
            "Epoch 6/250\n",
            " - 6s - loss: 0.0010 - val_loss: 0.0012\n",
            "Epoch 7/250\n",
            " - 6s - loss: 9.7664e-04 - val_loss: 0.0011\n",
            "Epoch 8/250\n",
            " - 6s - loss: 9.2993e-04 - val_loss: 0.0011\n",
            "Epoch 9/250\n",
            " - 6s - loss: 9.0760e-04 - val_loss: 0.0010\n",
            "Epoch 10/250\n",
            " - 6s - loss: 8.8151e-04 - val_loss: 9.6267e-04\n",
            "Epoch 11/250\n",
            " - 6s - loss: 8.7005e-04 - val_loss: 9.7475e-04\n",
            "Epoch 12/250\n",
            " - 6s - loss: 8.6207e-04 - val_loss: 9.3355e-04\n",
            "Epoch 13/250\n",
            " - 6s - loss: 8.5650e-04 - val_loss: 9.3659e-04\n",
            "Epoch 14/250\n",
            " - 6s - loss: 8.4672e-04 - val_loss: 9.2914e-04\n",
            "Epoch 15/250\n",
            " - 6s - loss: 8.4035e-04 - val_loss: 9.3467e-04\n",
            "Epoch 16/250\n",
            " - 6s - loss: 8.3076e-04 - val_loss: 9.1751e-04\n",
            "Epoch 17/250\n",
            " - 6s - loss: 8.2611e-04 - val_loss: 8.6262e-04\n",
            "Epoch 18/250\n",
            " - 6s - loss: 8.1484e-04 - val_loss: 9.0862e-04\n",
            "Epoch 19/250\n",
            " - 6s - loss: 8.1134e-04 - val_loss: 9.3350e-04\n",
            "Epoch 20/250\n",
            " - 6s - loss: 8.0035e-04 - val_loss: 9.3596e-04\n",
            "Epoch 21/250\n",
            " - 6s - loss: 7.9007e-04 - val_loss: 9.1606e-04\n",
            "Epoch 22/250\n",
            " - 6s - loss: 7.8083e-04 - val_loss: 8.1865e-04\n",
            "Epoch 23/250\n",
            " - 6s - loss: 7.3725e-04 - val_loss: 7.5220e-04\n",
            "Epoch 24/250\n",
            " - 6s - loss: 6.7154e-04 - val_loss: 7.2173e-04\n",
            "Epoch 25/250\n",
            " - 6s - loss: 6.0978e-04 - val_loss: 6.6944e-04\n",
            "Epoch 26/250\n",
            " - 6s - loss: 5.0487e-04 - val_loss: 5.0906e-04\n",
            "Epoch 27/250\n",
            " - 6s - loss: 3.4918e-04 - val_loss: 3.2580e-04\n",
            "Epoch 28/250\n",
            " - 6s - loss: 2.7019e-04 - val_loss: 3.1722e-04\n",
            "Epoch 29/250\n",
            " - 6s - loss: 2.4034e-04 - val_loss: 2.6157e-04\n",
            "Epoch 30/250\n",
            " - 6s - loss: 2.2967e-04 - val_loss: 2.4522e-04\n",
            "Epoch 31/250\n",
            " - 6s - loss: 2.1137e-04 - val_loss: 2.5731e-04\n",
            "Epoch 32/250\n",
            " - 6s - loss: 2.0978e-04 - val_loss: 2.3194e-04\n",
            "Epoch 33/250\n",
            " - 6s - loss: 2.0852e-04 - val_loss: 2.3046e-04\n",
            "Epoch 34/250\n",
            " - 6s - loss: 2.0272e-04 - val_loss: 3.8316e-04\n",
            "Epoch 35/250\n",
            " - 6s - loss: 2.0348e-04 - val_loss: 2.5143e-04\n",
            "Epoch 36/250\n",
            " - 6s - loss: 1.9869e-04 - val_loss: 2.6365e-04\n",
            "Epoch 37/250\n",
            " - 6s - loss: 1.9886e-04 - val_loss: 2.7641e-04\n",
            "Epoch 38/250\n",
            " - 6s - loss: 1.9341e-04 - val_loss: 2.6371e-04\n",
            "Epoch 39/250\n",
            " - 6s - loss: 1.9512e-04 - val_loss: 2.4066e-04\n",
            "Epoch 40/250\n",
            " - 6s - loss: 1.9264e-04 - val_loss: 2.6373e-04\n",
            "Epoch 41/250\n",
            " - 6s - loss: 1.9193e-04 - val_loss: 2.5372e-04\n",
            "Epoch 42/250\n",
            " - 6s - loss: 1.9490e-04 - val_loss: 2.2007e-04\n",
            "Epoch 43/250\n",
            " - 6s - loss: 1.9701e-04 - val_loss: 2.2019e-04\n",
            "Epoch 44/250\n",
            " - 6s - loss: 1.9045e-04 - val_loss: 2.3891e-04\n",
            "Epoch 45/250\n",
            " - 6s - loss: 1.8801e-04 - val_loss: 2.1867e-04\n",
            "Epoch 46/250\n",
            " - 6s - loss: 1.8717e-04 - val_loss: 2.1515e-04\n",
            "Epoch 47/250\n",
            " - 6s - loss: 1.8467e-04 - val_loss: 2.4281e-04\n",
            "Epoch 48/250\n",
            " - 6s - loss: 1.8804e-04 - val_loss: 2.1565e-04\n",
            "Epoch 49/250\n",
            " - 6s - loss: 1.8899e-04 - val_loss: 2.3765e-04\n",
            "Epoch 50/250\n",
            " - 6s - loss: 1.8778e-04 - val_loss: 2.1379e-04\n",
            "Epoch 51/250\n",
            " - 6s - loss: 1.9090e-04 - val_loss: 2.3094e-04\n",
            "Epoch 52/250\n",
            " - 6s - loss: 1.8322e-04 - val_loss: 2.5479e-04\n",
            "Epoch 53/250\n",
            " - 6s - loss: 1.8634e-04 - val_loss: 2.5962e-04\n",
            "Epoch 54/250\n",
            " - 6s - loss: 1.8655e-04 - val_loss: 3.0944e-04\n",
            "Epoch 55/250\n",
            " - 6s - loss: 1.9278e-04 - val_loss: 2.5113e-04\n",
            "Epoch 56/250\n",
            " - 6s - loss: 1.8628e-04 - val_loss: 2.3754e-04\n",
            "Epoch 57/250\n",
            " - 6s - loss: 1.8481e-04 - val_loss: 2.2529e-04\n",
            "Epoch 58/250\n",
            " - 6s - loss: 1.8163e-04 - val_loss: 2.4361e-04\n",
            "Epoch 59/250\n",
            " - 6s - loss: 1.8852e-04 - val_loss: 2.1731e-04\n",
            "Epoch 60/250\n",
            " - 6s - loss: 1.8245e-04 - val_loss: 2.7569e-04\n",
            "Epoch 61/250\n",
            " - 6s - loss: 1.8144e-04 - val_loss: 3.0498e-04\n",
            "Epoch 62/250\n",
            " - 6s - loss: 1.8439e-04 - val_loss: 2.2873e-04\n",
            "Epoch 63/250\n",
            " - 6s - loss: 1.8170e-04 - val_loss: 2.0552e-04\n",
            "Epoch 64/250\n",
            " - 6s - loss: 1.8422e-04 - val_loss: 2.2197e-04\n",
            "Epoch 65/250\n",
            " - 6s - loss: 1.8341e-04 - val_loss: 2.1579e-04\n",
            "Epoch 66/250\n",
            " - 6s - loss: 1.8037e-04 - val_loss: 2.0372e-04\n",
            "Epoch 67/250\n",
            " - 6s - loss: 1.8409e-04 - val_loss: 2.6189e-04\n",
            "Epoch 68/250\n",
            " - 6s - loss: 1.7877e-04 - val_loss: 2.3580e-04\n",
            "Epoch 69/250\n",
            " - 6s - loss: 1.7590e-04 - val_loss: 2.1707e-04\n",
            "Epoch 70/250\n",
            " - 6s - loss: 1.8085e-04 - val_loss: 2.3010e-04\n",
            "Epoch 71/250\n",
            " - 6s - loss: 1.7982e-04 - val_loss: 2.5415e-04\n",
            "Epoch 72/250\n",
            " - 6s - loss: 1.7835e-04 - val_loss: 2.1461e-04\n",
            "Epoch 73/250\n",
            " - 6s - loss: 1.8297e-04 - val_loss: 2.4800e-04\n",
            "Epoch 74/250\n",
            " - 6s - loss: 1.7487e-04 - val_loss: 2.7494e-04\n",
            "Epoch 75/250\n",
            " - 6s - loss: 1.7680e-04 - val_loss: 2.2453e-04\n",
            "Epoch 76/250\n",
            " - 6s - loss: 1.7429e-04 - val_loss: 2.4107e-04\n",
            "Epoch 77/250\n",
            " - 6s - loss: 1.7519e-04 - val_loss: 2.0524e-04\n",
            "Epoch 78/250\n",
            " - 6s - loss: 1.8083e-04 - val_loss: 2.0916e-04\n",
            "Epoch 79/250\n",
            " - 6s - loss: 1.7552e-04 - val_loss: 2.4719e-04\n",
            "Epoch 80/250\n",
            " - 6s - loss: 1.7764e-04 - val_loss: 2.1267e-04\n",
            "Epoch 81/250\n",
            " - 6s - loss: 1.7676e-04 - val_loss: 2.1117e-04\n",
            "Epoch 82/250\n",
            " - 6s - loss: 1.7457e-04 - val_loss: 2.3322e-04\n",
            "Epoch 83/250\n",
            " - 6s - loss: 1.7499e-04 - val_loss: 2.5521e-04\n",
            "Epoch 84/250\n",
            " - 6s - loss: 1.7857e-04 - val_loss: 2.4893e-04\n",
            "Epoch 85/250\n",
            " - 6s - loss: 1.7810e-04 - val_loss: 2.4365e-04\n",
            "Epoch 86/250\n",
            " - 6s - loss: 1.7727e-04 - val_loss: 2.3248e-04\n",
            "Epoch 87/250\n",
            " - 6s - loss: 1.7648e-04 - val_loss: 2.1871e-04\n",
            "Epoch 88/250\n",
            " - 6s - loss: 1.7254e-04 - val_loss: 2.5531e-04\n",
            "Epoch 89/250\n",
            " - 6s - loss: 1.7573e-04 - val_loss: 2.2722e-04\n",
            "Epoch 90/250\n",
            " - 6s - loss: 1.7444e-04 - val_loss: 2.3080e-04\n",
            "Epoch 91/250\n",
            " - 6s - loss: 1.7401e-04 - val_loss: 2.3620e-04\n",
            "Epoch 92/250\n",
            " - 6s - loss: 1.7414e-04 - val_loss: 2.5244e-04\n",
            "Epoch 93/250\n",
            " - 6s - loss: 1.7227e-04 - val_loss: 2.1817e-04\n",
            "Epoch 94/250\n",
            " - 6s - loss: 1.7893e-04 - val_loss: 2.2545e-04\n",
            "Epoch 95/250\n",
            " - 6s - loss: 1.7539e-04 - val_loss: 2.3264e-04\n",
            "Epoch 96/250\n",
            " - 6s - loss: 1.6929e-04 - val_loss: 2.0739e-04\n",
            "Epoch 97/250\n",
            " - 6s - loss: 1.6956e-04 - val_loss: 3.0525e-04\n",
            "Epoch 98/250\n",
            " - 6s - loss: 1.7079e-04 - val_loss: 1.9445e-04\n",
            "Epoch 99/250\n",
            " - 6s - loss: 1.7472e-04 - val_loss: 2.1052e-04\n",
            "Epoch 100/250\n",
            " - 6s - loss: 1.7353e-04 - val_loss: 2.4611e-04\n",
            "Epoch 101/250\n",
            " - 6s - loss: 1.7313e-04 - val_loss: 2.0358e-04\n",
            "Epoch 102/250\n",
            " - 6s - loss: 1.6823e-04 - val_loss: 2.4365e-04\n",
            "Epoch 103/250\n",
            " - 6s - loss: 1.6951e-04 - val_loss: 2.1829e-04\n",
            "Epoch 104/250\n",
            " - 6s - loss: 1.7076e-04 - val_loss: 2.4292e-04\n",
            "Epoch 105/250\n",
            " - 6s - loss: 1.7174e-04 - val_loss: 2.4008e-04\n",
            "Epoch 106/250\n",
            " - 6s - loss: 1.7278e-04 - val_loss: 2.0464e-04\n",
            "Epoch 107/250\n",
            " - 6s - loss: 1.6951e-04 - val_loss: 2.4182e-04\n",
            "Epoch 108/250\n",
            " - 6s - loss: 1.7082e-04 - val_loss: 2.2641e-04\n",
            "Epoch 109/250\n",
            " - 6s - loss: 1.7266e-04 - val_loss: 1.8967e-04\n",
            "Epoch 110/250\n",
            " - 6s - loss: 1.7187e-04 - val_loss: 2.0049e-04\n",
            "Epoch 111/250\n",
            " - 6s - loss: 1.6911e-04 - val_loss: 2.4423e-04\n",
            "Epoch 112/250\n",
            " - 6s - loss: 1.6929e-04 - val_loss: 2.1335e-04\n",
            "Epoch 113/250\n",
            " - 6s - loss: 1.7041e-04 - val_loss: 1.9032e-04\n",
            "Epoch 114/250\n",
            " - 6s - loss: 1.6640e-04 - val_loss: 2.4201e-04\n",
            "Epoch 115/250\n",
            " - 6s - loss: 1.7232e-04 - val_loss: 2.2036e-04\n",
            "Epoch 116/250\n",
            " - 6s - loss: 1.7353e-04 - val_loss: 2.2401e-04\n",
            "Epoch 117/250\n",
            " - 6s - loss: 1.6884e-04 - val_loss: 2.1535e-04\n",
            "Epoch 118/250\n",
            " - 6s - loss: 1.6841e-04 - val_loss: 2.1247e-04\n",
            "Epoch 119/250\n",
            " - 6s - loss: 1.6933e-04 - val_loss: 1.9798e-04\n",
            "Epoch 120/250\n",
            " - 6s - loss: 1.6773e-04 - val_loss: 2.1626e-04\n",
            "Epoch 121/250\n",
            " - 6s - loss: 1.6562e-04 - val_loss: 1.9844e-04\n",
            "Epoch 122/250\n",
            " - 6s - loss: 1.6508e-04 - val_loss: 2.2457e-04\n",
            "Epoch 123/250\n",
            " - 6s - loss: 1.6858e-04 - val_loss: 2.0921e-04\n",
            "Epoch 124/250\n",
            " - 6s - loss: 1.6971e-04 - val_loss: 2.1020e-04\n",
            "Epoch 125/250\n",
            " - 6s - loss: 1.6706e-04 - val_loss: 2.0267e-04\n",
            "Epoch 126/250\n",
            " - 6s - loss: 1.6732e-04 - val_loss: 2.6997e-04\n",
            "Epoch 127/250\n",
            " - 6s - loss: 1.6959e-04 - val_loss: 2.3336e-04\n",
            "Epoch 128/250\n",
            " - 6s - loss: 1.6577e-04 - val_loss: 2.1429e-04\n",
            "Epoch 129/250\n",
            " - 6s - loss: 1.7106e-04 - val_loss: 2.0607e-04\n",
            "Epoch 130/250\n",
            " - 6s - loss: 1.6171e-04 - val_loss: 2.5774e-04\n",
            "Epoch 131/250\n",
            " - 6s - loss: 1.6983e-04 - val_loss: 2.2336e-04\n",
            "Epoch 132/250\n",
            " - 6s - loss: 1.6878e-04 - val_loss: 2.5885e-04\n",
            "Epoch 133/250\n",
            " - 6s - loss: 1.6614e-04 - val_loss: 2.3070e-04\n",
            "Epoch 134/250\n",
            " - 6s - loss: 1.6811e-04 - val_loss: 2.8333e-04\n",
            "Epoch 135/250\n",
            " - 6s - loss: 1.7007e-04 - val_loss: 2.2264e-04\n",
            "Epoch 136/250\n",
            " - 6s - loss: 1.6679e-04 - val_loss: 2.0597e-04\n",
            "Epoch 137/250\n",
            " - 6s - loss: 1.7009e-04 - val_loss: 2.3883e-04\n",
            "Epoch 138/250\n",
            " - 6s - loss: 1.6984e-04 - val_loss: 2.1179e-04\n",
            "Epoch 139/250\n",
            " - 6s - loss: 1.6588e-04 - val_loss: 2.1329e-04\n",
            "Epoch 140/250\n",
            " - 6s - loss: 1.6517e-04 - val_loss: 2.5589e-04\n",
            "Epoch 141/250\n",
            " - 6s - loss: 1.6580e-04 - val_loss: 1.9768e-04\n",
            "Epoch 142/250\n",
            " - 6s - loss: 1.6520e-04 - val_loss: 2.7867e-04\n",
            "Epoch 143/250\n",
            " - 6s - loss: 1.6773e-04 - val_loss: 2.2383e-04\n",
            "Epoch 144/250\n",
            " - 6s - loss: 1.6629e-04 - val_loss: 2.1143e-04\n",
            "Epoch 145/250\n",
            " - 6s - loss: 1.6631e-04 - val_loss: 2.3798e-04\n",
            "Epoch 146/250\n",
            " - 6s - loss: 1.6510e-04 - val_loss: 2.2607e-04\n",
            "Epoch 147/250\n",
            " - 6s - loss: 1.6525e-04 - val_loss: 2.1926e-04\n",
            "Epoch 148/250\n",
            " - 6s - loss: 1.6777e-04 - val_loss: 1.9969e-04\n",
            "Epoch 149/250\n",
            " - 6s - loss: 1.6389e-04 - val_loss: 2.0440e-04\n",
            "Epoch 150/250\n",
            " - 6s - loss: 1.6328e-04 - val_loss: 1.9207e-04\n",
            "Epoch 151/250\n",
            " - 6s - loss: 1.6803e-04 - val_loss: 1.9804e-04\n",
            "Epoch 152/250\n",
            " - 6s - loss: 1.6717e-04 - val_loss: 1.9144e-04\n",
            "Epoch 153/250\n",
            " - 6s - loss: 1.6063e-04 - val_loss: 2.2082e-04\n",
            "Epoch 154/250\n",
            " - 6s - loss: 1.6326e-04 - val_loss: 1.8664e-04\n",
            "Epoch 155/250\n",
            " - 6s - loss: 1.6722e-04 - val_loss: 2.0049e-04\n",
            "Epoch 156/250\n",
            " - 6s - loss: 1.6323e-04 - val_loss: 2.2480e-04\n",
            "Epoch 157/250\n",
            " - 6s - loss: 1.6566e-04 - val_loss: 2.2972e-04\n",
            "Epoch 158/250\n",
            " - 6s - loss: 1.6527e-04 - val_loss: 2.3896e-04\n",
            "Epoch 159/250\n",
            " - 6s - loss: 1.6664e-04 - val_loss: 2.0089e-04\n",
            "Epoch 160/250\n",
            " - 6s - loss: 1.6334e-04 - val_loss: 2.1976e-04\n",
            "Epoch 161/250\n",
            " - 6s - loss: 1.6389e-04 - val_loss: 2.1224e-04\n",
            "Epoch 162/250\n",
            " - 6s - loss: 1.6375e-04 - val_loss: 2.4676e-04\n",
            "Epoch 163/250\n",
            " - 6s - loss: 1.6668e-04 - val_loss: 2.3295e-04\n",
            "Epoch 164/250\n",
            " - 6s - loss: 1.6639e-04 - val_loss: 2.0335e-04\n",
            "Epoch 165/250\n",
            " - 6s - loss: 1.6211e-04 - val_loss: 1.9593e-04\n",
            "Epoch 166/250\n",
            " - 6s - loss: 1.6898e-04 - val_loss: 2.2479e-04\n",
            "Epoch 167/250\n",
            " - 6s - loss: 1.6540e-04 - val_loss: 2.0762e-04\n",
            "Epoch 168/250\n",
            " - 6s - loss: 1.6555e-04 - val_loss: 1.8991e-04\n",
            "Epoch 169/250\n",
            " - 6s - loss: 1.6319e-04 - val_loss: 2.1428e-04\n",
            "Epoch 170/250\n",
            " - 6s - loss: 1.6142e-04 - val_loss: 1.9892e-04\n",
            "Epoch 171/250\n",
            " - 6s - loss: 1.6341e-04 - val_loss: 2.1433e-04\n",
            "Epoch 172/250\n",
            " - 6s - loss: 1.6768e-04 - val_loss: 2.0885e-04\n",
            "Epoch 173/250\n",
            " - 6s - loss: 1.6359e-04 - val_loss: 2.2353e-04\n",
            "Epoch 174/250\n",
            " - 6s - loss: 1.6334e-04 - val_loss: 2.6657e-04\n",
            "Epoch 175/250\n",
            " - 6s - loss: 1.6264e-04 - val_loss: 2.1283e-04\n",
            "Epoch 176/250\n",
            " - 6s - loss: 1.5789e-04 - val_loss: 1.9271e-04\n",
            "Epoch 177/250\n",
            " - 6s - loss: 1.6212e-04 - val_loss: 2.2117e-04\n",
            "Epoch 178/250\n",
            " - 6s - loss: 1.6131e-04 - val_loss: 1.9886e-04\n",
            "Epoch 179/250\n",
            " - 6s - loss: 1.6379e-04 - val_loss: 2.0458e-04\n",
            "Epoch 180/250\n",
            " - 6s - loss: 1.6525e-04 - val_loss: 2.0189e-04\n",
            "Epoch 181/250\n",
            " - 6s - loss: 1.6235e-04 - val_loss: 1.9929e-04\n",
            "Epoch 182/250\n",
            " - 6s - loss: 1.6050e-04 - val_loss: 2.0295e-04\n",
            "Epoch 183/250\n",
            " - 6s - loss: 1.6477e-04 - val_loss: 2.0300e-04\n",
            "Epoch 184/250\n",
            " - 6s - loss: 1.6332e-04 - val_loss: 2.0119e-04\n",
            "Epoch 185/250\n",
            " - 6s - loss: 1.6237e-04 - val_loss: 2.1873e-04\n",
            "Epoch 186/250\n",
            " - 6s - loss: 1.6039e-04 - val_loss: 2.2039e-04\n",
            "Epoch 187/250\n",
            " - 6s - loss: 1.6161e-04 - val_loss: 2.1500e-04\n",
            "Epoch 188/250\n",
            " - 6s - loss: 1.6465e-04 - val_loss: 1.9105e-04\n",
            "Epoch 189/250\n",
            " - 6s - loss: 1.6300e-04 - val_loss: 2.3677e-04\n",
            "Epoch 190/250\n",
            " - 6s - loss: 1.6204e-04 - val_loss: 2.3196e-04\n",
            "Epoch 191/250\n",
            " - 6s - loss: 1.6313e-04 - val_loss: 2.1032e-04\n",
            "Epoch 192/250\n",
            " - 6s - loss: 1.6022e-04 - val_loss: 1.8903e-04\n",
            "Epoch 193/250\n",
            " - 6s - loss: 1.6083e-04 - val_loss: 2.2951e-04\n",
            "Epoch 194/250\n",
            " - 6s - loss: 1.5860e-04 - val_loss: 2.2399e-04\n",
            "Epoch 195/250\n",
            " - 6s - loss: 1.6462e-04 - val_loss: 2.0026e-04\n",
            "Epoch 196/250\n",
            " - 6s - loss: 1.5977e-04 - val_loss: 1.9945e-04\n",
            "Epoch 197/250\n",
            " - 6s - loss: 1.6002e-04 - val_loss: 1.9417e-04\n",
            "Epoch 198/250\n",
            " - 6s - loss: 1.6202e-04 - val_loss: 2.0279e-04\n",
            "Epoch 199/250\n",
            " - 6s - loss: 1.5877e-04 - val_loss: 1.9472e-04\n",
            "Epoch 200/250\n",
            " - 6s - loss: 1.6276e-04 - val_loss: 2.2232e-04\n",
            "Epoch 201/250\n",
            " - 6s - loss: 1.5981e-04 - val_loss: 2.0983e-04\n",
            "Epoch 202/250\n",
            " - 6s - loss: 1.6164e-04 - val_loss: 2.0739e-04\n",
            "Epoch 203/250\n",
            " - 6s - loss: 1.5871e-04 - val_loss: 2.6348e-04\n",
            "Epoch 204/250\n",
            " - 6s - loss: 1.6086e-04 - val_loss: 2.0521e-04\n",
            "14813/14813 [==============================] - 0s 15us/step\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.00019220800964338237"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 104
        },
        {
          "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": "iJc82JNX_SKs",
        "colab_type": "code",
        "outputId": "a7f8a192-23a5-4f7c-a58c-2c98ccb7d127",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        }
      },
      "source": [
        "plot_history(history)"
      ],
      "execution_count": 110,
      "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": "Y3GHvnnVoOVV",
        "colab_type": "code",
        "outputId": "1748e0fb-3c4a-47b3-e2ed-9422ce7e94e3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "pred_y = model.predict(test_x)\n",
        "score = r2_score(test_y, pred_y)\n",
        "print(f'Test r2 score: {score}')\n"
      ],
      "execution_count": 118,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Test r2 score: 0.993490265129513\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}