GNN_S&P500.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation, concatenate\n",
    "from keras.optimizers import RMSprop, SGD, Adam\n",
    "from keras.constraints import non_neg\n",
    "from keras.layers import Multiply\n",
    "from keras.layers import dot\n",
    "from keras import layers"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Data"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cut Test Data row\n",
    "euro : 68,023\n",
    "s&p : 162,908"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "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>Unnamed: 0</th>\n",
       "      <th>underlying</th>\n",
       "      <th>underlying_last</th>\n",
       "      <th>expiration</th>\n",
       "      <th>quotedate</th>\n",
       "      <th>strike</th>\n",
       "      <th>price</th>\n",
       "      <th>impliedvol</th>\n",
       "      <th>delta</th>\n",
       "      <th>gamma</th>\n",
       "      <th>theta</th>\n",
       "      <th>vega</th>\n",
       "      <th>interest rate</th>\n",
       "      <th>moneyness</th>\n",
       "      <th>tau</th>\n",
       "      <th>moneyness_class</th>\n",
       "      <th>tau_class</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2726.62</td>\n",
       "      <td>2018-07-20</td>\n",
       "      <td>2018-07-02</td>\n",
       "      <td>800</td>\n",
       "      <td>1924.25</td>\n",
       "      <td>0.0996</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-18.6218</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02</td>\n",
       "      <td>3.408275</td>\n",
       "      <td>0.075397</td>\n",
       "      <td>itm</td>\n",
       "      <td>shortterm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2726.62</td>\n",
       "      <td>2018-07-20</td>\n",
       "      <td>2018-07-02</td>\n",
       "      <td>900</td>\n",
       "      <td>1824.40</td>\n",
       "      <td>0.0996</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-20.9495</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02</td>\n",
       "      <td>3.029578</td>\n",
       "      <td>0.075397</td>\n",
       "      <td>itm</td>\n",
       "      <td>shortterm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2726.62</td>\n",
       "      <td>2018-07-20</td>\n",
       "      <td>2018-07-02</td>\n",
       "      <td>1000</td>\n",
       "      <td>1724.55</td>\n",
       "      <td>0.0996</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-23.2772</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02</td>\n",
       "      <td>2.726620</td>\n",
       "      <td>0.075397</td>\n",
       "      <td>itm</td>\n",
       "      <td>shortterm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2726.62</td>\n",
       "      <td>2018-07-20</td>\n",
       "      <td>2018-07-02</td>\n",
       "      <td>1050</td>\n",
       "      <td>1674.60</td>\n",
       "      <td>0.0996</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-24.4411</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02</td>\n",
       "      <td>2.596781</td>\n",
       "      <td>0.075397</td>\n",
       "      <td>itm</td>\n",
       "      <td>shortterm</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2726.62</td>\n",
       "      <td>2018-07-20</td>\n",
       "      <td>2018-07-02</td>\n",
       "      <td>1100</td>\n",
       "      <td>1624.65</td>\n",
       "      <td>0.0996</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-25.6049</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.02</td>\n",
       "      <td>2.478745</td>\n",
       "      <td>0.075397</td>\n",
       "      <td>itm</td>\n",
       "      <td>shortterm</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0 underlying  underlying_last  expiration   quotedate  strike  \\\n",
       "0           0        SPX          2726.62  2018-07-20  2018-07-02     800   \n",
       "1           1        SPX          2726.62  2018-07-20  2018-07-02     900   \n",
       "2           2        SPX          2726.62  2018-07-20  2018-07-02    1000   \n",
       "3           3        SPX          2726.62  2018-07-20  2018-07-02    1050   \n",
       "4           4        SPX          2726.62  2018-07-20  2018-07-02    1100   \n",
       "\n",
       "     price  impliedvol  delta  gamma    theta  vega  interest rate  moneyness  \\\n",
       "0  1924.25      0.0996    1.0    0.0 -18.6218   0.0           0.02   3.408275   \n",
       "1  1824.40      0.0996    1.0    0.0 -20.9495   0.0           0.02   3.029578   \n",
       "2  1724.55      0.0996    1.0    0.0 -23.2772   0.0           0.02   2.726620   \n",
       "3  1674.60      0.0996    1.0    0.0 -24.4411   0.0           0.02   2.596781   \n",
       "4  1624.65      0.0996    1.0    0.0 -25.6049   0.0           0.02   2.478745   \n",
       "\n",
       "        tau moneyness_class  tau_class  \n",
       "0  0.075397             itm  shortterm  \n",
       "1  0.075397             itm  shortterm  \n",
       "2  0.075397             itm  shortterm  \n",
       "3  0.075397             itm  shortterm  \n",
       "4  0.075397             itm  shortterm  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# S&P 500\n",
    "call_option = pd.read_csv(\"snp_revised_data.csv\")\n",
    "call_option.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# inverse moneyness\n",
    "call_option['inverse_moneyness']=1/call_option['moneyness']\n",
    "\n",
    "# transform market price: y=c*exp(rt)/St\n",
    "call_option['transform']=call_option['price']*np.exp(call_option['interest rate']*call_option['tau'])/call_option['underlying_last']\n",
    "\n",
    "# # create moneyness_class for call option\n",
    "# call_option['moneyness_class'] = np.where(call_option['moneyness']>1.05,'itm','atm')\n",
    "# call_option['moneyness_class'] = np.where(call_option['moneyness']<0.97,'otm',call_option['moneyness_class'])\n",
    "\n",
    "# # create tau_class for call option\n",
    "# call_option['tau_class'] = np.where(call_option['tau']>0.2,'longterm','midterm')\n",
    "# call_option['tau_class'] = np.where(call_option['tau']<0.1,'shortterm',call_option['tau_class'])\n",
    "\n",
    "#call_option"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train, validation, test\n",
    "data_train = call_option.iloc[:int(len(call_option)*0.7),:]\n",
    "data_validation = call_option.iloc[int(len(call_option)*0.7):162908,:]\n",
    "data_test= call_option.iloc[162908:,:]        "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras import backend as K\n",
    "def root_mean_squared_error(y_true, y_pred):\n",
    "        return K.sqrt(K.mean(K.square(y_pred - y_true), axis=-1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 9 - model\n",
    "# left multi model \n",
    "# model1\n",
    "model_1_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_1_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_1_left_dense = Dense(5, activation=\"softplus\")(model_1_left_input)\n",
    "model_1_right_dense = Dense(5, activation=\"sigmoid\")(model_1_right_input)\n",
    "\n",
    "model_1_multi = layers.multiply([model_1_left_dense, model_1_right_dense])\n",
    "model_1_out = Dense(1, activation=\"linear\", use_bias=False)(model_1_multi)\n",
    "model_1 = keras.models.Model(inputs=[model_1_left_input, model_1_right_input], outputs=model_1_out)\n",
    "\n",
    "# model 2\n",
    "model_2_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_2_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_2_left_dense = Dense(5, activation=\"softplus\")(model_2_left_input)\n",
    "model_2_right_dense = Dense(5, activation=\"sigmoid\")(model_2_right_input)\n",
    "\n",
    "model_2_multi = layers.multiply([model_2_left_dense, model_2_right_dense])\n",
    "model_2_out = Dense(1, activation=\"linear\", use_bias=False)(model_2_multi)\n",
    "model_2 = keras.models.Model(inputs=[model_2_left_input,model_2_right_input], outputs = model_2_out)\n",
    "\n",
    "# model3\n",
    "model_3_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_3_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_3_left_dense = Dense(5, activation=\"softplus\")(model_3_left_input)\n",
    "model_3_right_dense = Dense(5, activation=\"sigmoid\")(model_3_right_input)\n",
    "\n",
    "model_3_multi = layers.multiply([model_3_left_dense, model_3_right_dense])\n",
    "model_3_out = Dense(1, activation=\"linear\", use_bias=False)(model_3_multi)\n",
    "model_3 = keras.models.Model(inputs=[model_3_left_input,model_3_right_input], outputs = model_3_out)\n",
    "\n",
    "# model4\n",
    "model_4_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_4_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_4_left_dense = Dense(5, activation=\"softplus\")(model_4_left_input)\n",
    "model_4_right_dense = Dense(5, activation=\"sigmoid\")(model_4_right_input)\n",
    "\n",
    "model_4_multi = layers.multiply([model_4_left_dense, model_4_right_dense])\n",
    "model_4_out = Dense(1, activation=\"linear\", use_bias=False)(model_4_multi)\n",
    "model_4 = keras.models.Model(inputs=[model_4_left_input,model_4_right_input], outputs = model_4_out)\n",
    "\n",
    "# model_5\n",
    "model_5_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_5_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_5_left_dense = Dense(5, activation=\"softplus\")(model_5_left_input)\n",
    "model_5_right_dense = Dense(5, activation=\"sigmoid\")(model_5_right_input)\n",
    "\n",
    "model_5_multi = layers.multiply([model_5_left_dense, model_5_right_dense])\n",
    "model_5_out = Dense(1, activation=\"linear\", use_bias=False)(model_5_multi)\n",
    "model_5 = keras.models.Model(inputs=[model_5_left_input,model_5_right_input], outputs = model_5_out)\n",
    "\n",
    "# model_6\n",
    "model_6_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_6_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_6_left_dense = Dense(5, activation=\"softplus\")(model_6_left_input)\n",
    "model_6_right_dense = Dense(5, activation=\"sigmoid\")(model_6_right_input)\n",
    "\n",
    "model_6_multi = layers.multiply([model_6_left_dense, model_6_right_dense])\n",
    "model_6_out = Dense(1, activation=\"linear\", use_bias=False)(model_6_multi)\n",
    "model_6 = keras.models.Model(inputs=[model_6_left_input,model_6_right_input], outputs = model_6_out)\n",
    "\n",
    "# model_7\n",
    "model_7_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_7_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_7_left_dense = Dense(5, activation=\"softplus\")(model_7_left_input)\n",
    "model_7_right_dense = Dense(5, activation=\"sigmoid\")(model_7_right_input)\n",
    "\n",
    "model_7_multi = layers.multiply([model_7_left_dense, model_7_right_dense])\n",
    "model_7_out = Dense(1, activation=\"linear\", use_bias=False)(model_7_multi)\n",
    "model_7 = keras.models.Model(inputs=[model_7_left_input,model_7_right_input], outputs = model_7_out)\n",
    "\n",
    "# model_8\n",
    "model_8_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_8_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_8_left_dense = Dense(5, activation=\"softplus\")(model_8_left_input)\n",
    "model_8_right_dense = Dense(5, activation=\"sigmoid\")(model_8_right_input)\n",
    "\n",
    "model_8_multi = layers.multiply([model_8_left_dense, model_8_right_dense])\n",
    "model_8_out = Dense(1, activation=\"linear\", use_bias=False)(model_8_multi)\n",
    "model_8 = keras.models.Model(inputs=[model_8_left_input,model_8_right_input], outputs = model_8_out)\n",
    "\n",
    "# model_9\n",
    "model_9_left_input = keras.layers.Input(shape=(1, ))\n",
    "model_9_right_input = keras.layers.Input(shape=(1, ))\n",
    "model_9_left_dense = Dense(5, activation=\"softplus\")(model_9_left_input)\n",
    "model_9_right_dense = Dense(5, activation=\"sigmoid\")(model_9_right_input)\n",
    "\n",
    "model_9_multi = layers.multiply([model_9_left_dense, model_9_right_dense])\n",
    "model_9_out = Dense(1, activation=\"linear\", use_bias=False)(model_9_multi)\n",
    "model_9 = keras.models.Model(inputs=[model_9_left_input,model_9_right_input], outputs = model_9_out)\n",
    "\n",
    "\n",
    "model_concat = concatenate([model_1_out, model_2_out, model_3_out,\n",
    "                           model_4_out, model_5_out, model_6_out,\n",
    "                           model_7_out, model_8_out, model_9_out,])\n",
    "\n",
    "# right model\n",
    "model_right_input = keras.layers.Input(shape=(2, ))\n",
    "model_right_1 = Dense(5, activation=\"sigmoid\")(model_right_input)\n",
    "model_right_2 = Dense(9, activation=\"softplus\")(model_right_1)\n",
    "\n",
    "# final model\n",
    "model_final = dot([model_concat,model_right_2],axes = 1,normalize=False)\n",
    "model = keras.models.Model(inputs=[model_1_left_input, model_1_right_input,\n",
    "                                   model_2_left_input, model_2_right_input,\n",
    "                                   model_3_left_input, model_3_right_input,\n",
    "                                   model_4_left_input, model_4_right_input,\n",
    "                                   model_5_left_input, model_5_right_input,\n",
    "                                   model_6_left_input, model_6_right_input,\n",
    "                                   model_7_left_input, model_7_right_input,\n",
    "                                   model_8_left_input, model_8_right_input,\n",
    "                                   model_9_left_input, model_9_right_input, model_right_input],\n",
    "                           outputs = model_final)\n",
    "# keras.utils.plot_model(model, show_shapes=True)\n",
    "\n",
    "# model compile\n",
    "model.compile(loss='mean_squared_error',\n",
    "                      optimizer = Adam(), #RMSprop(),SGD(),Adam()\n",
    "                      metrics=['mape', root_mean_squared_error])              "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "#keras.utils.plot_model(model, show_shapes=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### fit data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# normalize\n",
    "sc1 = StandardScaler() # fit training feature1\n",
    "sc2 = StandardScaler() # fit training feature2\n",
    "sc3 = StandardScaler() # fit validation feature1 and/or target (shared without reuse conflict)\n",
    "sc4 = StandardScaler() # fit validation feature2,can reuse\n",
    "sc5 = StandardScaler() # fit training target, inverse transform for test prediction\n",
    "sc7 = StandardScaler()\n",
    "sc8 = StandardScaler()  # can reuse\n",
    "\n",
    "# early stopping\n",
    "es = keras.callbacks.EarlyStopping(monitor='val_loss',min_delta=0,patience=30, verbose=2, mode='auto') \n",
    "\n",
    "# inverse_moneyness - fit\n",
    "X1_train = sc1.fit_transform(-data_train[['inverse_moneyness']])\n",
    "X2_train = sc2.fit_transform(data_train[['tau']])\n",
    "X1_validation = sc3.fit_transform(-data_validation[['inverse_moneyness']])\n",
    "X2_validation = sc4.fit_transform(data_validation[['tau']])\n",
    "right_train = sc7.fit_transform(data_train[['inverse_moneyness','tau']])\n",
    "right_validation = sc8.fit_transform(data_validation[['inverse_moneyness','tau']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 143586 samples, validate on 19322 samples\n",
      "Epoch 1/100\n",
      " - 6s - loss: 0.1011 - mape: 79.6556 - root_mean_squared_error: 0.1052 - val_loss: 0.0044 - val_mape: 11.4144 - val_root_mean_squared_error: 0.0472\n",
      "Epoch 2/100\n",
      " - 5s - loss: 0.0019 - mape: 36.6905 - root_mean_squared_error: 0.0312 - val_loss: 0.0023 - val_mape: 18.4159 - val_root_mean_squared_error: 0.0358\n",
      "Epoch 3/100\n",
      " - 5s - loss: 9.9885e-04 - mape: 25.6794 - root_mean_squared_error: 0.0232 - val_loss: 0.0019 - val_mape: 18.6571 - val_root_mean_squared_error: 0.0330\n",
      "Epoch 4/100\n",
      " - 5s - loss: 7.1865e-04 - mape: 18.7504 - root_mean_squared_error: 0.0197 - val_loss: 0.0019 - val_mape: 23.9972 - val_root_mean_squared_error: 0.0346\n",
      "Epoch 5/100\n",
      " - 5s - loss: 5.4008e-04 - mape: 16.6114 - root_mean_squared_error: 0.0173 - val_loss: 0.0015 - val_mape: 31.6051 - val_root_mean_squared_error: 0.0329\n",
      "Epoch 6/100\n",
      " - 5s - loss: 4.0704e-04 - mape: 13.2649 - root_mean_squared_error: 0.0150 - val_loss: 0.0016 - val_mape: 32.0152 - val_root_mean_squared_error: 0.0343\n",
      "Epoch 7/100\n",
      " - 5s - loss: 3.4019e-04 - mape: 11.8694 - root_mean_squared_error: 0.0136 - val_loss: 0.0014 - val_mape: 29.7167 - val_root_mean_squared_error: 0.0321\n",
      "Epoch 8/100\n",
      " - 5s - loss: 3.1436e-04 - mape: 10.6391 - root_mean_squared_error: 0.0130 - val_loss: 0.0016 - val_mape: 27.8373 - val_root_mean_squared_error: 0.0326\n",
      "Epoch 9/100\n",
      " - 5s - loss: 3.0274e-04 - mape: 12.7254 - root_mean_squared_error: 0.0128 - val_loss: 0.0015 - val_mape: 23.5340 - val_root_mean_squared_error: 0.0318\n",
      "Epoch 10/100\n",
      " - 5s - loss: 2.9008e-04 - mape: 10.2994 - root_mean_squared_error: 0.0124 - val_loss: 0.0015 - val_mape: 32.2384 - val_root_mean_squared_error: 0.0322\n",
      "Epoch 11/100\n",
      " - 5s - loss: 2.8249e-04 - mape: 10.0022 - root_mean_squared_error: 0.0122 - val_loss: 0.0015 - val_mape: 25.7396 - val_root_mean_squared_error: 0.0323\n",
      "Epoch 12/100\n",
      " - 5s - loss: 2.8112e-04 - mape: 10.6571 - root_mean_squared_error: 0.0122 - val_loss: 0.0014 - val_mape: 26.0501 - val_root_mean_squared_error: 0.0308\n",
      "Epoch 13/100\n",
      " - 5s - loss: 2.7502e-04 - mape: 9.6351 - root_mean_squared_error: 0.0120 - val_loss: 0.0014 - val_mape: 23.5444 - val_root_mean_squared_error: 0.0312\n",
      "Epoch 14/100\n",
      " - 5s - loss: 2.7085e-04 - mape: 10.8053 - root_mean_squared_error: 0.0119 - val_loss: 0.0015 - val_mape: 26.7854 - val_root_mean_squared_error: 0.0324\n",
      "Epoch 15/100\n",
      " - 5s - loss: 2.6953e-04 - mape: 10.5903 - root_mean_squared_error: 0.0119 - val_loss: 0.0016 - val_mape: 28.2108 - val_root_mean_squared_error: 0.0333\n",
      "Epoch 16/100\n",
      " - 5s - loss: 2.6617e-04 - mape: 10.1859 - root_mean_squared_error: 0.0118 - val_loss: 0.0016 - val_mape: 27.2386 - val_root_mean_squared_error: 0.0334\n",
      "Epoch 17/100\n",
      " - 5s - loss: 2.6401e-04 - mape: 11.7178 - root_mean_squared_error: 0.0117 - val_loss: 0.0015 - val_mape: 27.3568 - val_root_mean_squared_error: 0.0325\n",
      "Epoch 18/100\n",
      " - 5s - loss: 2.6195e-04 - mape: 11.2036 - root_mean_squared_error: 0.0117 - val_loss: 0.0015 - val_mape: 27.1495 - val_root_mean_squared_error: 0.0321\n",
      "Epoch 19/100\n",
      " - 5s - loss: 2.6170e-04 - mape: 9.5297 - root_mean_squared_error: 0.0117 - val_loss: 0.0016 - val_mape: 25.2504 - val_root_mean_squared_error: 0.0331\n",
      "Epoch 20/100\n",
      " - 5s - loss: 2.5843e-04 - mape: 9.7404 - root_mean_squared_error: 0.0115 - val_loss: 0.0014 - val_mape: 20.5555 - val_root_mean_squared_error: 0.0299\n",
      "Epoch 21/100\n",
      " - 5s - loss: 2.5779e-04 - mape: 9.0776 - root_mean_squared_error: 0.0115 - val_loss: 0.0017 - val_mape: 24.8409 - val_root_mean_squared_error: 0.0331\n",
      "Epoch 22/100\n",
      " - 5s - loss: 2.5714e-04 - mape: 8.9755 - root_mean_squared_error: 0.0115 - val_loss: 0.0014 - val_mape: 25.3898 - val_root_mean_squared_error: 0.0313\n",
      "Epoch 23/100\n",
      " - 5s - loss: 2.5485e-04 - mape: 10.7249 - root_mean_squared_error: 0.0114 - val_loss: 0.0016 - val_mape: 23.1245 - val_root_mean_squared_error: 0.0327\n",
      "Epoch 24/100\n",
      " - 5s - loss: 2.5598e-04 - mape: 10.0504 - root_mean_squared_error: 0.0115 - val_loss: 0.0015 - val_mape: 27.0954 - val_root_mean_squared_error: 0.0319\n",
      "Epoch 25/100\n",
      " - 5s - loss: 2.5427e-04 - mape: 9.6737 - root_mean_squared_error: 0.0114 - val_loss: 0.0015 - val_mape: 25.2765 - val_root_mean_squared_error: 0.0316\n",
      "Epoch 26/100\n",
      " - 5s - loss: 2.5286e-04 - mape: 10.7801 - root_mean_squared_error: 0.0114 - val_loss: 0.0016 - val_mape: 27.7760 - val_root_mean_squared_error: 0.0327\n",
      "Epoch 27/100\n",
      " - 5s - loss: 2.5124e-04 - mape: 9.9474 - root_mean_squared_error: 0.0113 - val_loss: 0.0017 - val_mape: 27.2699 - val_root_mean_squared_error: 0.0336\n",
      "Epoch 28/100\n",
      " - 5s - loss: 2.5115e-04 - mape: 10.0265 - root_mean_squared_error: 0.0113 - val_loss: 0.0016 - val_mape: 24.1019 - val_root_mean_squared_error: 0.0325\n",
      "Epoch 29/100\n",
      " - 5s - loss: 2.5049e-04 - mape: 10.2946 - root_mean_squared_error: 0.0113 - val_loss: 0.0016 - val_mape: 25.5147 - val_root_mean_squared_error: 0.0332\n",
      "Epoch 30/100\n",
      " - 5s - loss: 2.4783e-04 - mape: 10.9029 - root_mean_squared_error: 0.0112 - val_loss: 0.0016 - val_mape: 26.8192 - val_root_mean_squared_error: 0.0335\n",
      "Epoch 31/100\n",
      " - 5s - loss: 2.4953e-04 - mape: 10.2958 - root_mean_squared_error: 0.0113 - val_loss: 0.0015 - val_mape: 27.9266 - val_root_mean_squared_error: 0.0326\n",
      "Epoch 32/100\n",
      " - 5s - loss: 2.4829e-04 - mape: 10.2188 - root_mean_squared_error: 0.0112 - val_loss: 0.0017 - val_mape: 28.5504 - val_root_mean_squared_error: 0.0340\n",
      "Epoch 33/100\n",
      " - 6s - loss: 2.4734e-04 - mape: 10.4956 - root_mean_squared_error: 0.0112 - val_loss: 0.0016 - val_mape: 31.0065 - val_root_mean_squared_error: 0.0335\n",
      "Epoch 34/100\n",
      " - 6s - loss: 2.4718e-04 - mape: 9.8689 - root_mean_squared_error: 0.0112 - val_loss: 0.0016 - val_mape: 26.3746 - val_root_mean_squared_error: 0.0330\n",
      "Epoch 35/100\n",
      " - 6s - loss: 2.4615e-04 - mape: 10.7247 - root_mean_squared_error: 0.0111 - val_loss: 0.0015 - val_mape: 27.1118 - val_root_mean_squared_error: 0.0327\n",
      "Epoch 36/100\n",
      " - 6s - loss: 2.4502e-04 - mape: 10.1631 - root_mean_squared_error: 0.0111 - val_loss: 0.0015 - val_mape: 24.8885 - val_root_mean_squared_error: 0.0322\n",
      "Epoch 37/100\n",
      " - 6s - loss: 2.4454e-04 - mape: 9.7734 - root_mean_squared_error: 0.0111 - val_loss: 0.0016 - val_mape: 25.2990 - val_root_mean_squared_error: 0.0329\n",
      "Epoch 38/100\n",
      " - 6s - loss: 2.4435e-04 - mape: 9.4006 - root_mean_squared_error: 0.0111 - val_loss: 0.0017 - val_mape: 27.9417 - val_root_mean_squared_error: 0.0339\n",
      "Epoch 39/100\n",
      " - 6s - loss: 2.4347e-04 - mape: 9.9635 - root_mean_squared_error: 0.0111 - val_loss: 0.0015 - val_mape: 25.9899 - val_root_mean_squared_error: 0.0328\n",
      "Epoch 40/100\n",
      " - 6s - loss: 2.4360e-04 - mape: 9.8907 - root_mean_squared_error: 0.0111 - val_loss: 0.0016 - val_mape: 24.7495 - val_root_mean_squared_error: 0.0333\n",
      "Epoch 41/100\n",
      " - 7s - loss: 2.4118e-04 - mape: 9.7678 - root_mean_squared_error: 0.0110 - val_loss: 0.0014 - val_mape: 24.1036 - val_root_mean_squared_error: 0.0307\n",
      "Epoch 42/100\n",
      " - 6s - loss: 2.4171e-04 - mape: 10.7121 - root_mean_squared_error: 0.0110 - val_loss: 0.0015 - val_mape: 29.0319 - val_root_mean_squared_error: 0.0326\n",
      "Epoch 43/100\n",
      " - 6s - loss: 2.4286e-04 - mape: 9.9443 - root_mean_squared_error: 0.0110 - val_loss: 0.0016 - val_mape: 27.5876 - val_root_mean_squared_error: 0.0335\n",
      "Epoch 44/100\n",
      " - 7s - loss: 2.4112e-04 - mape: 10.4402 - root_mean_squared_error: 0.0110 - val_loss: 0.0015 - val_mape: 22.6521 - val_root_mean_squared_error: 0.0314\n",
      "Epoch 45/100\n",
      " - 6s - loss: 2.4002e-04 - mape: 10.2155 - root_mean_squared_error: 0.0109 - val_loss: 0.0016 - val_mape: 26.2251 - val_root_mean_squared_error: 0.0329\n",
      "Epoch 46/100\n",
      " - 7s - loss: 2.3923e-04 - mape: 9.4121 - root_mean_squared_error: 0.0109 - val_loss: 0.0015 - val_mape: 25.3271 - val_root_mean_squared_error: 0.0323\n",
      "Epoch 47/100\n",
      " - 6s - loss: 2.3977e-04 - mape: 11.0694 - root_mean_squared_error: 0.0109 - val_loss: 0.0014 - val_mape: 23.4455 - val_root_mean_squared_error: 0.0314\n",
      "Epoch 48/100\n",
      " - 7s - loss: 2.3938e-04 - mape: 9.2032 - root_mean_squared_error: 0.0109 - val_loss: 0.0016 - val_mape: 25.9918 - val_root_mean_squared_error: 0.0330\n",
      "Epoch 49/100\n",
      " - 7s - loss: 2.4014e-04 - mape: 8.9582 - root_mean_squared_error: 0.0110 - val_loss: 0.0015 - val_mape: 26.7394 - val_root_mean_squared_error: 0.0323\n",
      "Epoch 50/100\n",
      " - 6s - loss: 2.3879e-04 - mape: 10.9427 - root_mean_squared_error: 0.0109 - val_loss: 0.0015 - val_mape: 28.4184 - val_root_mean_squared_error: 0.0324\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00050: early stopping\n"
     ]
    }
   ],
   "source": [
    "# no kernerl constraint + inverse_moneyness\n",
    "history = model.fit([X1_train, X2_train, X1_train, X2_train, X1_train, X2_train,\n",
    "                     X1_train, X2_train, X1_train, X2_train, X1_train, X2_train,\n",
    "                     X1_train, X2_train, X1_train, X2_train, X1_train, X2_train,\n",
    "                     right_train], sc5.fit_transform(data_train[['transform']]),\n",
    "                    batch_size=50, epochs=100, verbose=2,\n",
    "                    validation_data=([X1_validation, X2_validation, X1_validation, X2_validation, X1_validation, X2_validation,\n",
    "                                      X1_validation, X2_validation, X1_validation, X2_validation, X1_validation, X2_validation,\n",
    "                                      X1_validation, X2_validation, X1_validation, X2_validation, X1_validation, X2_validation,\n",
    "                                      right_validation],\n",
    "                                     sc3.fit_transform(data_validation[['transform']])), callbacks=[es]) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "performance = []\n",
    "years_data = pd.DataFrame()\n",
    "\n",
    "# inverse_moneyness predict\n",
    "X1_test = sc1.transform(-data_test[['inverse_moneyness']])\n",
    "X2_test = sc2.transform(data_test[['tau']])\n",
    "right_test = sc7.transform(data_test[['inverse_moneyness','tau']])\n",
    "\n",
    "prediction = model.predict([X1_test, X2_test, X1_test, X2_test, X1_test, X2_test,\n",
    "                            X1_test, X2_test, X1_test, X2_test, X1_test, X2_test,\n",
    "                            X1_test, X2_test, X1_test, X2_test, X1_test, X2_test,\n",
    "                            right_test])\n",
    "\n",
    "# percentage mean-squared-error\n",
    "data_test_copy = data_test.copy()\n",
    "data_test_copy['temp'] = sc5.inverse_transform(prediction)  # inverse transform\n",
    "data_test_copy['GNN_price'] = data_test_copy['temp']*data_test['underlying_last']*np.exp(-data_test['interest rate']*data_test['tau'])\n",
    "data_test_copy['GNN_square_error'] = ((data_test['price'] - data_test_copy['GNN_price'])/data_test['price'])**2\n",
    "data_test_copy = data_test_copy.drop(['temp'],1)\n",
    "\n",
    "performance.append(data_test_copy['GNN_square_error'].mean())     \n",
    "years_data = years_data.append(data_test_copy) \n",
    "\n",
    "# GNN Price Clipping\n",
    "years_data['GNN_price_clip'] = np.clip(years_data['GNN_price'],0,max(years_data['GNN_price']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "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>Unnamed: 0</th>\n",
       "      <th>underlying</th>\n",
       "      <th>underlying_last</th>\n",
       "      <th>expiration</th>\n",
       "      <th>quotedate</th>\n",
       "      <th>strike</th>\n",
       "      <th>price</th>\n",
       "      <th>impliedvol</th>\n",
       "      <th>delta</th>\n",
       "      <th>gamma</th>\n",
       "      <th>...</th>\n",
       "      <th>interest rate</th>\n",
       "      <th>moneyness</th>\n",
       "      <th>tau</th>\n",
       "      <th>moneyness_class</th>\n",
       "      <th>tau_class</th>\n",
       "      <th>inverse_moneyness</th>\n",
       "      <th>transform</th>\n",
       "      <th>GNN_price</th>\n",
       "      <th>GNN_square_error</th>\n",
       "      <th>GNN_price_clip</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>162908</th>\n",
       "      <td>162908</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2641.88</td>\n",
       "      <td>2018-12-21</td>\n",
       "      <td>2018-11-20</td>\n",
       "      <td>100</td>\n",
       "      <td>2534.40</td>\n",
       "      <td>0.1735</td>\n",
       "      <td>0.9980</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0225</td>\n",
       "      <td>26.418800</td>\n",
       "      <td>0.126984</td>\n",
       "      <td>itm</td>\n",
       "      <td>midterm</td>\n",
       "      <td>0.037852</td>\n",
       "      <td>0.962062</td>\n",
       "      <td>2545.553923</td>\n",
       "      <td>0.000019</td>\n",
       "      <td>2545.553923</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>162909</th>\n",
       "      <td>162909</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2641.88</td>\n",
       "      <td>2018-12-21</td>\n",
       "      <td>2018-11-20</td>\n",
       "      <td>200</td>\n",
       "      <td>2434.75</td>\n",
       "      <td>0.1735</td>\n",
       "      <td>0.9980</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0225</td>\n",
       "      <td>13.209400</td>\n",
       "      <td>0.126984</td>\n",
       "      <td>itm</td>\n",
       "      <td>midterm</td>\n",
       "      <td>0.075704</td>\n",
       "      <td>0.924234</td>\n",
       "      <td>2444.038908</td>\n",
       "      <td>0.000015</td>\n",
       "      <td>2444.038908</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>162910</th>\n",
       "      <td>162910</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2641.88</td>\n",
       "      <td>2018-12-21</td>\n",
       "      <td>2018-11-20</td>\n",
       "      <td>250</td>\n",
       "      <td>2384.70</td>\n",
       "      <td>0.1735</td>\n",
       "      <td>0.9980</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0225</td>\n",
       "      <td>10.567520</td>\n",
       "      <td>0.126984</td>\n",
       "      <td>itm</td>\n",
       "      <td>midterm</td>\n",
       "      <td>0.094630</td>\n",
       "      <td>0.905235</td>\n",
       "      <td>2393.410941</td>\n",
       "      <td>0.000013</td>\n",
       "      <td>2393.410941</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>162911</th>\n",
       "      <td>162911</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2641.88</td>\n",
       "      <td>2018-12-21</td>\n",
       "      <td>2018-11-20</td>\n",
       "      <td>300</td>\n",
       "      <td>2334.85</td>\n",
       "      <td>0.1735</td>\n",
       "      <td>0.9980</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0225</td>\n",
       "      <td>8.806267</td>\n",
       "      <td>0.126984</td>\n",
       "      <td>itm</td>\n",
       "      <td>midterm</td>\n",
       "      <td>0.113555</td>\n",
       "      <td>0.886312</td>\n",
       "      <td>2342.869178</td>\n",
       "      <td>0.000012</td>\n",
       "      <td>2342.869178</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>162912</th>\n",
       "      <td>162912</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2641.88</td>\n",
       "      <td>2018-12-21</td>\n",
       "      <td>2018-11-20</td>\n",
       "      <td>400</td>\n",
       "      <td>2235.10</td>\n",
       "      <td>0.1735</td>\n",
       "      <td>0.9980</td>\n",
       "      <td>0.0000</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0225</td>\n",
       "      <td>6.604700</td>\n",
       "      <td>0.126984</td>\n",
       "      <td>itm</td>\n",
       "      <td>midterm</td>\n",
       "      <td>0.151407</td>\n",
       "      <td>0.848447</td>\n",
       "      <td>2242.041435</td>\n",
       "      <td>0.000010</td>\n",
       "      <td>2242.041435</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",
       "      <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>205118</th>\n",
       "      <td>205118</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2506.85</td>\n",
       "      <td>2021-12-17</td>\n",
       "      <td>2018-12-31</td>\n",
       "      <td>3200</td>\n",
       "      <td>71.55</td>\n",
       "      <td>0.1432</td>\n",
       "      <td>0.2231</td>\n",
       "      <td>0.0005</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0250</td>\n",
       "      <td>0.783391</td>\n",
       "      <td>4.297619</td>\n",
       "      <td>otm</td>\n",
       "      <td>longterm</td>\n",
       "      <td>1.276502</td>\n",
       "      <td>0.031779</td>\n",
       "      <td>52.648984</td>\n",
       "      <td>0.069783</td>\n",
       "      <td>52.648984</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>205119</th>\n",
       "      <td>205119</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2506.85</td>\n",
       "      <td>2021-12-17</td>\n",
       "      <td>2018-12-31</td>\n",
       "      <td>3300</td>\n",
       "      <td>54.80</td>\n",
       "      <td>0.1407</td>\n",
       "      <td>0.1830</td>\n",
       "      <td>0.0004</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0250</td>\n",
       "      <td>0.759652</td>\n",
       "      <td>4.297619</td>\n",
       "      <td>otm</td>\n",
       "      <td>longterm</td>\n",
       "      <td>1.316393</td>\n",
       "      <td>0.024340</td>\n",
       "      <td>37.071626</td>\n",
       "      <td>0.104659</td>\n",
       "      <td>37.071626</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>205120</th>\n",
       "      <td>205120</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2506.85</td>\n",
       "      <td>2021-12-17</td>\n",
       "      <td>2018-12-31</td>\n",
       "      <td>3400</td>\n",
       "      <td>42.10</td>\n",
       "      <td>0.1392</td>\n",
       "      <td>0.1494</td>\n",
       "      <td>0.0004</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0250</td>\n",
       "      <td>0.737309</td>\n",
       "      <td>4.297619</td>\n",
       "      <td>otm</td>\n",
       "      <td>longterm</td>\n",
       "      <td>1.356284</td>\n",
       "      <td>0.018699</td>\n",
       "      <td>25.659992</td>\n",
       "      <td>0.152489</td>\n",
       "      <td>25.659992</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>205121</th>\n",
       "      <td>205121</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2506.85</td>\n",
       "      <td>2021-12-17</td>\n",
       "      <td>2018-12-31</td>\n",
       "      <td>3500</td>\n",
       "      <td>32.20</td>\n",
       "      <td>0.1380</td>\n",
       "      <td>0.1209</td>\n",
       "      <td>0.0003</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0250</td>\n",
       "      <td>0.716243</td>\n",
       "      <td>4.297619</td>\n",
       "      <td>otm</td>\n",
       "      <td>longterm</td>\n",
       "      <td>1.396174</td>\n",
       "      <td>0.014302</td>\n",
       "      <td>17.856886</td>\n",
       "      <td>0.198415</td>\n",
       "      <td>17.856886</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>205122</th>\n",
       "      <td>205122</td>\n",
       "      <td>SPX</td>\n",
       "      <td>2506.85</td>\n",
       "      <td>2021-12-17</td>\n",
       "      <td>2018-12-31</td>\n",
       "      <td>3600</td>\n",
       "      <td>25.00</td>\n",
       "      <td>0.1372</td>\n",
       "      <td>0.0974</td>\n",
       "      <td>0.0003</td>\n",
       "      <td>...</td>\n",
       "      <td>0.0250</td>\n",
       "      <td>0.696347</td>\n",
       "      <td>4.297619</td>\n",
       "      <td>otm</td>\n",
       "      <td>longterm</td>\n",
       "      <td>1.436065</td>\n",
       "      <td>0.011104</td>\n",
       "      <td>12.961860</td>\n",
       "      <td>0.231867</td>\n",
       "      <td>12.961860</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>42215 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        Unnamed: 0 underlying  underlying_last  expiration   quotedate  \\\n",
       "162908      162908        SPX          2641.88  2018-12-21  2018-11-20   \n",
       "162909      162909        SPX          2641.88  2018-12-21  2018-11-20   \n",
       "162910      162910        SPX          2641.88  2018-12-21  2018-11-20   \n",
       "162911      162911        SPX          2641.88  2018-12-21  2018-11-20   \n",
       "162912      162912        SPX          2641.88  2018-12-21  2018-11-20   \n",
       "...            ...        ...              ...         ...         ...   \n",
       "205118      205118        SPX          2506.85  2021-12-17  2018-12-31   \n",
       "205119      205119        SPX          2506.85  2021-12-17  2018-12-31   \n",
       "205120      205120        SPX          2506.85  2021-12-17  2018-12-31   \n",
       "205121      205121        SPX          2506.85  2021-12-17  2018-12-31   \n",
       "205122      205122        SPX          2506.85  2021-12-17  2018-12-31   \n",
       "\n",
       "        strike    price  impliedvol   delta   gamma  ...  interest rate  \\\n",
       "162908     100  2534.40      0.1735  0.9980  0.0000  ...         0.0225   \n",
       "162909     200  2434.75      0.1735  0.9980  0.0000  ...         0.0225   \n",
       "162910     250  2384.70      0.1735  0.9980  0.0000  ...         0.0225   \n",
       "162911     300  2334.85      0.1735  0.9980  0.0000  ...         0.0225   \n",
       "162912     400  2235.10      0.1735  0.9980  0.0000  ...         0.0225   \n",
       "...        ...      ...         ...     ...     ...  ...            ...   \n",
       "205118    3200    71.55      0.1432  0.2231  0.0005  ...         0.0250   \n",
       "205119    3300    54.80      0.1407  0.1830  0.0004  ...         0.0250   \n",
       "205120    3400    42.10      0.1392  0.1494  0.0004  ...         0.0250   \n",
       "205121    3500    32.20      0.1380  0.1209  0.0003  ...         0.0250   \n",
       "205122    3600    25.00      0.1372  0.0974  0.0003  ...         0.0250   \n",
       "\n",
       "        moneyness       tau  moneyness_class  tau_class inverse_moneyness  \\\n",
       "162908  26.418800  0.126984              itm    midterm          0.037852   \n",
       "162909  13.209400  0.126984              itm    midterm          0.075704   \n",
       "162910  10.567520  0.126984              itm    midterm          0.094630   \n",
       "162911   8.806267  0.126984              itm    midterm          0.113555   \n",
       "162912   6.604700  0.126984              itm    midterm          0.151407   \n",
       "...           ...       ...              ...        ...               ...   \n",
       "205118   0.783391  4.297619              otm   longterm          1.276502   \n",
       "205119   0.759652  4.297619              otm   longterm          1.316393   \n",
       "205120   0.737309  4.297619              otm   longterm          1.356284   \n",
       "205121   0.716243  4.297619              otm   longterm          1.396174   \n",
       "205122   0.696347  4.297619              otm   longterm          1.436065   \n",
       "\n",
       "       transform    GNN_price  GNN_square_error  GNN_price_clip  \n",
       "162908  0.962062  2545.553923          0.000019     2545.553923  \n",
       "162909  0.924234  2444.038908          0.000015     2444.038908  \n",
       "162910  0.905235  2393.410941          0.000013     2393.410941  \n",
       "162911  0.886312  2342.869178          0.000012     2342.869178  \n",
       "162912  0.848447  2242.041435          0.000010     2242.041435  \n",
       "...          ...          ...               ...             ...  \n",
       "205118  0.031779    52.648984          0.069783       52.648984  \n",
       "205119  0.024340    37.071626          0.104659       37.071626  \n",
       "205120  0.018699    25.659992          0.152489       25.659992  \n",
       "205121  0.014302    17.856886          0.198415       17.856886  \n",
       "205122  0.011104    12.961860          0.231867       12.961860  \n",
       "\n",
       "[42215 rows x 22 columns]"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "years_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "years_data.to_csv('./GNN_snp_predict.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE ==  13.0899211008666\n",
      "MAPE ==  0.4620056523845975\n",
      "MSE ==  353.9745836546551\n",
      "RMSE ==  18.814212278345725\n"
     ]
    }
   ],
   "source": [
    "y_true = years_data['price']\n",
    "y_pred = years_data['GNN_price']\n",
    "\n",
    "# MAE\n",
    "def MAE(y_true, y_pred): \n",
    "    return np.mean(np.abs((y_true - y_pred)))\n",
    "\n",
    "print(\"MAE == \", MAE(y_true, y_pred))\n",
    "\n",
    "# MAPE\n",
    "def MAPE(y_true, y_pred): \n",
    "    return np.mean(np.abs((y_true - y_pred) / y_true))\n",
    "\n",
    "print(\"MAPE == \", MAPE(y_true, y_pred))\n",
    "\n",
    "# MSE\n",
    "def MSE(y_true, y_pred):\n",
    "    return np.mean(np.square((y_true - y_pred)))\n",
    "\n",
    "print(\"MSE == \", MSE(y_true, y_pred))\n",
    "\n",
    "#RMSE\n",
    "print(\"RMSE == \", np.sqrt(MSE(y_true, y_pred)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MAE ==  12.950215485210194\n",
      "MAPE ==  0.41924975119876695\n",
      "MSE ==  351.2461148538932\n",
      "RMSE ==  18.741561163731618\n"
     ]
    }
   ],
   "source": [
    "# Clip GNN_price\n",
    "y_true = years_data['price']\n",
    "y_pred = years_data['GNN_price_clip']\n",
    "\n",
    "# MAE\n",
    "def MAE(y_true, y_pred): \n",
    "    return np.mean(np.abs((y_true - y_pred)))\n",
    "\n",
    "print(\"MAE == \", MAE(y_true, y_pred))\n",
    "\n",
    "# MAPE\n",
    "def MAPE(y_true, y_pred): \n",
    "    return np.mean(np.abs((y_true - y_pred) / y_true))\n",
    "\n",
    "print(\"MAPE == \", MAPE(y_true, y_pred))\n",
    "\n",
    "# MSE\n",
    "def MSE(y_true, y_pred):\n",
    "    return np.mean(np.square((y_true - y_pred)))\n",
    "\n",
    "print(\"MSE == \", MSE(y_true, y_pred))\n",
    "\n",
    "#RMSE\n",
    "print(\"RMSE == \", np.sqrt(MSE(y_true, y_pred)))\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}