proj_lstm_forecaster.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "66ff5158",
   "metadata": {},
   "outputs": [],
   "source": [
    "# imports\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "plt.rcParams[\"figure.figsize\"] = (20,10)\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.model_selection import train_test_split\n",
    "from keras.layers.recurrent import LSTM\n",
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Dropout, Activation\n",
    "from sklearn import preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "8ccdf350",
   "metadata": {},
   "outputs": [],
   "source": [
    "# going to use tsla\n",
    "fn ='./proj_data_TSLA.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "e8bfefb3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(26276, 6)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# just loading the data here, if you want to see how it is gathered / graphs check Proj_pull_data \n",
    "df = pd.read_csv(fn, index_col=0)\n",
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "fb53e0ec",
   "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>open</th>\n",
       "      <th>high</th>\n",
       "      <th>low</th>\n",
       "      <th>close</th>\n",
       "      <th>volume</th>\n",
       "      <th>datetime</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>843.45</td>\n",
       "      <td>846.00</td>\n",
       "      <td>843.45</td>\n",
       "      <td>844.94</td>\n",
       "      <td>2431</td>\n",
       "      <td>2021-10-18 08:00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>845.00</td>\n",
       "      <td>845.00</td>\n",
       "      <td>844.15</td>\n",
       "      <td>844.44</td>\n",
       "      <td>303</td>\n",
       "      <td>2021-10-18 08:01:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>846.00</td>\n",
       "      <td>846.00</td>\n",
       "      <td>846.00</td>\n",
       "      <td>846.00</td>\n",
       "      <td>300</td>\n",
       "      <td>2021-10-18 08:02:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>845.11</td>\n",
       "      <td>845.11</td>\n",
       "      <td>844.99</td>\n",
       "      <td>844.99</td>\n",
       "      <td>400</td>\n",
       "      <td>2021-10-18 08:03:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>845.61</td>\n",
       "      <td>845.61</td>\n",
       "      <td>845.00</td>\n",
       "      <td>845.00</td>\n",
       "      <td>918</td>\n",
       "      <td>2021-10-18 08:05:00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     open    high     low   close  volume             datetime\n",
       "0  843.45  846.00  843.45  844.94    2431  2021-10-18 08:00:00\n",
       "1  845.00  845.00  844.15  844.44     303  2021-10-18 08:01:00\n",
       "2  846.00  846.00  846.00  846.00     300  2021-10-18 08:02:00\n",
       "3  845.11  845.11  844.99  844.99     400  2021-10-18 08:03:00\n",
       "4  845.61  845.61  845.00  845.00     918  2021-10-18 08:05:00"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# data is minute data for about 46 days minus after hours\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "102a7c3d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# function for min-max normalization of stock\n",
    "def normalize_data(df):\n",
    "    min_max_scaler = MinMaxScaler()\n",
    "    df['open'] = min_max_scaler.fit_transform(df.open.values.reshape(-1,1))\n",
    "    df['high'] = min_max_scaler.fit_transform(df.high.values.reshape(-1,1))\n",
    "    df['low'] = min_max_scaler.fit_transform(df.low.values.reshape(-1,1))\n",
    "    df['close'] = min_max_scaler.fit_transform(df['close'].values.reshape(-1,1))\n",
    "    return df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "52b2bf5a-334d-4005-9168-5382ecf72c2e",
   "metadata": {},
   "source": [
    "### Creating Target Labels  \n",
    "The main idea is to calculate the percent change from the last high to the current high, this will be a percentage between -100 and 100. From there classify by the percent change as outlined below. After looking at the data there are very few values outside the range of -.5 to .5 \n",
    "\n",
    "The y is calculated as the percent change:  \n",
    "$$\n",
    "  Y=\\frac{(X_n-X_{n+1})}{X_n}*100\n",
    "$$\n",
    "From there the percent change is divided into 10 classes from -1 to 1 being:  \n",
    "[-100  -80  -60  -40  -20 0 20 40 60 80  100]  \n",
    "which correspond to class  \n",
    "[ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "7765d784-2d50-412d-aa92-b6c76179f906",
   "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>open</th>\n",
       "      <th>high</th>\n",
       "      <th>low</th>\n",
       "      <th>close</th>\n",
       "      <th>y</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.003845</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.007986</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.006326</td>\n",
       "      <td>0.002469</td>\n",
       "      <td>0.012564</td>\n",
       "      <td>0.003886</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.004118</td>\n",
       "      <td>0.000272</td>\n",
       "      <td>0.010065</td>\n",
       "      <td>0.001370</td>\n",
       "      <td>-1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.005358</td>\n",
       "      <td>0.001506</td>\n",
       "      <td>0.010090</td>\n",
       "      <td>0.001395</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.006276</td>\n",
       "      <td>0.002469</td>\n",
       "      <td>0.010090</td>\n",
       "      <td>0.001395</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       open      high       low     close    y\n",
       "1  0.003845  0.000000  0.007986  0.000000 -1.0\n",
       "2  0.006326  0.002469  0.012564  0.003886  1.0\n",
       "3  0.004118  0.000272  0.010065  0.001370 -1.0\n",
       "4  0.005358  0.001506  0.010090  0.001395  1.0\n",
       "5  0.006276  0.002469  0.010090  0.001395  1.0"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# scaler fits data to between 0 and 1 \n",
    "# todo can we normalize all test/train or need to separate?\n",
    "dfn = normalize_data(df)\n",
    "\n",
    "# now lets turn this into a dataset with a window of data and a forecast target\n",
    "# so the y value will be the 'next' minute's high value\n",
    "# we calculate the %change between the 'next' high and the 'last' high\n",
    "# y = (X - X\n",
    "# NOTE !! a lot of manual tuning went in here to find a good way to scale data so there are a reasonable number of classes that are close to same size ! \n",
    "dfn['y'] = (((dfn['high']-dfn['high'].shift(1))/dfn['high'])*100).clip(-1,1)\n",
    "# force everything to this interval \n",
    "dfn['y'] = (dfn['y']//.5)*.5\n",
    "# we will clip our values at between -20 and +20 because there are very very few values outside that range\n",
    "# note that is just for this dataset, for hour candles it would be much different\n",
    "dfn['y'] = dfn['y'].round(2)\n",
    "# we will also get rid of the other columns and clean up nans\n",
    "df2 = dfn[['open','high','low','close','y']]\n",
    "# there will be a na for the first row only\n",
    "df2 = df2.replace([np.inf, -np.inf], np.nan)\n",
    "df2 = df2.dropna()\n",
    "# integers become the classes\n",
    "\n",
    "#df2['y'] = df2['y'].astype(int)\n",
    "df2.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "e512bc6a-a7b8-491e-8a99-ca5b362b1e3b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-1.0    3997\n",
      "-0.5    9278\n",
      " 0.0    8971\n",
      " 0.5    2099\n",
      " 1.0    1930\n",
      "Name: y, dtype: int64\n",
      "Float64Index([-1.0, -0.5, 0.0, 0.5, 1.0], dtype='float64')\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0    3997\n",
       "1    9278\n",
       "2    8971\n",
       "3    2099\n",
       "4    1930\n",
       "Name: yc, dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# let's clean up our classes, instead of -5,5 0,10\n",
    "# new column 'yc' is Y Class\n",
    "df2['yc'] = df2['y']\n",
    "vc = df2['y'].value_counts(sort=False).sort_index(ascending=True)\n",
    "print(vc)\n",
    "keys = vc.keys()\n",
    "print(keys)\n",
    "df2['yc'].replace(keys, range(0,len(keys)), inplace=True)\n",
    "df2['yc'] = df2['yc'].astype(int)\n",
    "df2['yc'].value_counts().sort_index()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a3f8ffb2-4907-4dcd-82e7-e656778de177",
   "metadata": {},
   "source": [
    "### Understanding the class labels  \n",
    "Let's take a second to talk about the labels. Please note the labels have changed since I wrote this, this is just an example.   \n",
    "-0.24    2041  = 0 -> large negative  \n",
    "-0.16    2651  = 1 -> medium negative  \n",
    "-0.08    7787  = 2 -> small negative  \n",
    " 0.00    9102  = 3 -> small positive  \n",
    " 0.08    2136  = 4 -> medium positive  \n",
    " 0.16    2550  = 5 -> large positive  \n",
    " After much tweaking I was able to get 6 classes with a reasonable distrubution across all of them. So this way we don't have a horrible unbalanced dataset. So all of the values at 0.00 (9102) are between 0.0 and 0.08 which from a trading perspective would mean a very small amount of profit if you are long. Likewise for the 7787 values between 0.0 and -.08 except those values would be a small loss. We could further break it down into just two classes 'going up' or 'going down' but I think there is value in classifying as is meaning basically (from a long perspective) [big loss, medum loss, small loss, small gain, medium gain, big gain] which seems realistic.  \n",
    "From there we don't want long float numbers as 'classes' so we replace them with a simple integer number, we can refer back to this to understand what the classes are."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "df1b4447",
   "metadata": {},
   "outputs": [],
   "source": [
    "Xw = []\n",
    "Yw = []\n",
    "w=[]\n",
    "for i,r in df2.iterrows():\n",
    "    if i<21:\n",
    "        # add the first to the window\n",
    "        w.append(r[['open','high','low','close']].values)\n",
    "    else:\n",
    "        # append the other rows to the window [:20]\n",
    "        w.append(r[['open','high','low','close']].values)\n",
    "        w = w[1:] # pop off  the first entry\n",
    "        Xw.append(np.array(w))\n",
    "        # we will need to shift this later !!\n",
    "        Yw.append(r['yc'])\n",
    "    # for testing, break early\n",
    "    #if i > 100:\n",
    "    #    break\n",
    "        \n",
    "#push into numpy array\n",
    "Xw = np.array(Xw)\n",
    "Yw = np.array(Yw)\n",
    "# Shift Y it should be the answer for the 'next\n",
    "# just copy the first value and push into start, will shift all to the right\n",
    "# that's wrong it would make it more latent, not less\n",
    "#Yw = np.insert(Yw,0,Yw[0])\n",
    "# need to pop the first value instead\n",
    "Yw = Yw[2:]\n",
    "\n",
    "# strip off the X that have no y, since we only have y for the next we lose the first ones\n",
    "Xw = Xw[:Yw.shape[0]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b91ba524",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(26253, 20, 4)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Xw.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "ef08ce94",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(26253,)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Yw.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "f6d99aae",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[2.0, 4.0, 0.0, 0.0, 4.0, 0.0, 4.0, 0.0, 4.0, 4.0]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# just glance at our y 'windowed' values\n",
    "list(Yw[0:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "3f8869a5-08d9-46de-a344-0d9c32c1ebe9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[2, 4, 0, 0, 4, 0, 4, 0, 4, 4]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Convert to integers because that's what's expected for the sparce categorical\n",
    "Yw = Yw.astype(np.int32)\n",
    "list(Yw[0:10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "da8f67ac-62dc-4ac3-a4f1-cf203acc5c8e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X Windowed Shape: (26253, 20, 4)  Y Windowed Shape: (26253,)\n"
     ]
    }
   ],
   "source": [
    "print('X Windowed Shape:',Xw.shape,' Y Windowed Shape:',Yw.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "4cd1efe3-a7ad-42c0-8837-43c338b60426",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shapes: (19689, 20, 4) (6564, 20, 4) (19689,) (6564,)\n"
     ]
    }
   ],
   "source": [
    "#X_train ,X_test, y_train, y_test = train_test_split(Xw,tf.keras.utils.to_categorical(Yw), test_size=.25)\n",
    "X_train ,X_test, y_train, y_test = train_test_split(Xw,Yw, test_size=.25)\n",
    "print('Shapes:',X_train.shape ,X_test.shape, y_train.shape, y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "b704d7ab",
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_model(layers): # build a LSTM model\n",
    "    d = 0.3\n",
    "    model = Sequential()\n",
    "    \n",
    "    model.add(LSTM(64, input_shape=(layers[1], layers[0]), return_sequences=True))\n",
    "    model.add(Dropout(d)) # with probability of 0.3\n",
    "        \n",
    "    model.add(LSTM(64, input_shape=(layers[1], layers[0]), return_sequences=False))\n",
    "    model.add(Dropout(d)) # with probability of 0.3\n",
    "        \n",
    "    model.add(Dense(32,kernel_initializer=\"uniform\",activation='relu'))  \n",
    "          \n",
    "    model.add(Dense(layers[2],activation='relu'))\n",
    "    \n",
    "    # at 0.0001 acc max at 38%\n",
    "    # at 0.001 ?\n",
    "    model.compile(optimizer=tf.optimizers.Adam(learning_rate=0.001,\n",
    "                                       beta_1=0.9,\n",
    "                                       beta_2=0.999,\n",
    "                                       epsilon=1e-07,\n",
    "                                       amsgrad=False,\n",
    "                                       name='Adam'\n",
    "                                       ),\n",
    "                  loss='sparse_categorical_crossentropy', metrics=['accuracy'])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "b6e0cc45-2c9c-4a3d-ac14-54e4a6020b03",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm (LSTM)                  (None, 20, 64)            17664     \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 20, 64)            0         \n",
      "_________________________________________________________________\n",
      "lstm_1 (LSTM)                (None, 64)                33024     \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 5)                 165       \n",
      "=================================================================\n",
      "Total params: 52,933\n",
      "Trainable params: 52,933\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Epoch 1/20\n",
      "200/200 [==============================] - 99s 468ms/step - loss: 1.6611 - accuracy: 0.3634 - val_loss: 1.3563 - val_accuracy: 0.3756\n",
      "Epoch 2/20\n",
      "200/200 [==============================] - 94s 469ms/step - loss: 1.3667 - accuracy: 0.3807 - val_loss: 1.3525 - val_accuracy: 0.3779\n",
      "Epoch 3/20\n",
      "200/200 [==============================] - 99s 494ms/step - loss: 1.3554 - accuracy: 0.3774 - val_loss: 1.3509 - val_accuracy: 0.3743\n",
      "Epoch 4/20\n",
      "200/200 [==============================] - 92s 460ms/step - loss: 1.3631 - accuracy: 0.3815 - val_loss: 1.3502 - val_accuracy: 0.3756\n",
      "Epoch 5/20\n",
      "200/200 [==============================] - 93s 467ms/step - loss: 1.3856 - accuracy: 0.3663 - val_loss: 1.3672 - val_accuracy: 0.3545\n",
      "Epoch 6/20\n",
      "200/200 [==============================] - 91s 456ms/step - loss: 1.3613 - accuracy: 0.3736 - val_loss: 1.3525 - val_accuracy: 0.3735\n",
      "Epoch 7/20\n",
      "200/200 [==============================] - 92s 459ms/step - loss: 1.3604 - accuracy: 0.3716 - val_loss: 1.3502 - val_accuracy: 0.3740\n",
      "Epoch 8/20\n",
      "200/200 [==============================] - 94s 470ms/step - loss: 1.3569 - accuracy: 0.3791 - val_loss: 1.3498 - val_accuracy: 0.3768\n",
      "Epoch 9/20\n",
      "200/200 [==============================] - 95s 473ms/step - loss: 1.3571 - accuracy: 0.3819 - val_loss: 1.3475 - val_accuracy: 0.3702\n",
      "Epoch 10/20\n",
      "200/200 [==============================] - 95s 473ms/step - loss: 1.3539 - accuracy: 0.3811 - val_loss: 1.3464 - val_accuracy: 0.3674\n",
      "Epoch 11/20\n",
      "200/200 [==============================] - 91s 457ms/step - loss: 1.3557 - accuracy: 0.3790 - val_loss: 1.3473 - val_accuracy: 0.3723\n",
      "Epoch 12/20\n",
      "200/200 [==============================] - 92s 458ms/step - loss: 1.3517 - accuracy: 0.3787 - val_loss: 1.3441 - val_accuracy: 0.3743\n",
      "Epoch 13/20\n",
      "200/200 [==============================] - 91s 457ms/step - loss: 1.3525 - accuracy: 0.3826 - val_loss: 1.3473 - val_accuracy: 0.3738\n",
      "Epoch 14/20\n",
      "200/200 [==============================] - 92s 458ms/step - loss: 1.3533 - accuracy: 0.3821 - val_loss: 1.3470 - val_accuracy: 0.3733\n",
      "Epoch 15/20\n",
      "200/200 [==============================] - 92s 458ms/step - loss: 1.3543 - accuracy: 0.3813 - val_loss: 1.3474 - val_accuracy: 0.3758\n",
      "Epoch 16/20\n",
      "200/200 [==============================] - 91s 457ms/step - loss: 1.3546 - accuracy: 0.3814 - val_loss: 1.3451 - val_accuracy: 0.3761\n",
      "Epoch 17/20\n",
      "200/200 [==============================] - 91s 458ms/step - loss: 1.3563 - accuracy: 0.3790 - val_loss: 1.3460 - val_accuracy: 0.3766\n",
      "Epoch 18/20\n",
      "200/200 [==============================] - 92s 458ms/step - loss: 1.3468 - accuracy: 0.3808 - val_loss: 1.3487 - val_accuracy: 0.3776\n",
      "Epoch 19/20\n",
      "200/200 [==============================] - 92s 458ms/step - loss: 1.3604 - accuracy: 0.3801 - val_loss: 1.3560 - val_accuracy: 0.3756\n",
      "Epoch 20/20\n",
      "200/200 [==============================] - 91s 457ms/step - loss: 1.3486 - accuracy: 0.3891 - val_loss: 1.3449 - val_accuracy: 0.3763\n"
     ]
    }
   ],
   "source": [
    "model = build_model([4,20,Yw.max()+1])\n",
    "model.summary()\n",
    "hist = model.fit(X_train,y_train,validation_split=.2,batch_size=64,epochs=20,steps_per_epoch=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "5a5aedfb-3189-4cc8-8e0d-e191ccfa765a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "206/206 [==============================] - 12s 57ms/step - loss: 1.3524 - accuracy: 0.3844 ETA: 0s - loss: 1.3530 - accura\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[1.3524144887924194, 0.38436928391456604]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_test,y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "9ff07a98-efbc-4d04-a953-a83efb5809a8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot the loss, note since this is a time series problem accuracy does not make much sense\n",
    "#plt.plot(hist.history['loss'])\n",
    "plt.plot(hist.history['accuracy'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "66f3ac3f-023d-4c23-adfe-ffedee3f9188",
   "metadata": {},
   "source": [
    "### Save Point"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "ddbeee0f-5daa-4ecf-a902-b6ee2cc9f9f7",
   "metadata": {},
   "outputs": [],
   "source": [
    "# save off all data, so we can start again from here instead of redoing all above\n",
    "np.save('Xw.np',Xw)\n",
    "np.save('Yw.np',Yw)\n",
    "model.save('lstm_forecaster.h5')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1cf3ec17-d6b3-4891-a612-5e5c8c612ca8",
   "metadata": {},
   "source": [
    "### Conclusion  \n",
    "It is clear to see that this model fails to achive any decent results. Not only because the accuracy is so low, but also because there was little to no improvement from the starting accuracy. In other words improving from 35% accuracy to 37% accuracy is worse than improving from 5% to 25%, it shows there is no real improvement.  "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}