BTC conv net experiment.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "import quandl\n",
    "import numpy as np\n",
    "\n",
    "from keras import layers\n",
    "from keras.layers import Input, Dense, Activation, ZeroPadding2D, BatchNormalization, Flatten, Conv2D\n",
    "from keras.layers import AveragePooling2D, MaxPooling2D, Dropout, GlobalMaxPooling2D, GlobalAveragePooling2D\n",
    "from keras.models import Model\n",
    "from keras.preprocessing import image\n",
    "from keras.utils import layer_utils\n",
    "from keras.utils.data_utils import get_file\n",
    "from keras.applications.imagenet_utils import preprocess_input\n",
    "from keras.regularizers import l2\n",
    "\n",
    "import keras.backend as K\n",
    "from sklearn.model_selection import KFold\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#data = quandl.get('BCHARTS/BITFINEXUSD') -> OLD BUGGY DATASET (OBD)\n",
    "data = quandl.get('BITFINEX/BTCUSD')\n",
    "#data = quandl.get('BITFINEX/XRPBTC')\n",
    "#data = quandl.get('BITFINEX/ETHBTC')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Mid</th>\n",
       "      <th>Last</th>\n",
       "      <th>Bid</th>\n",
       "      <th>Ask</th>\n",
       "      <th>Volume</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2014-04-15</th>\n",
       "      <td>513.9000</td>\n",
       "      <td>452.00</td>\n",
       "      <td>504.23500</td>\n",
       "      <td>505.0000</td>\n",
       "      <td>503.5000</td>\n",
       "      <td>504.97</td>\n",
       "      <td>21013.584774</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2014-04-16</th>\n",
       "      <td>547.0000</td>\n",
       "      <td>495.00</td>\n",
       "      <td>537.50000</td>\n",
       "      <td>538.0000</td>\n",
       "      <td>537.0000</td>\n",
       "      <td>538.00</td>\n",
       "      <td>29633.358705</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2014-04-17</th>\n",
       "      <td>538.5000</td>\n",
       "      <td>486.10</td>\n",
       "      <td>507.02000</td>\n",
       "      <td>508.0000</td>\n",
       "      <td>506.0400</td>\n",
       "      <td>508.00</td>\n",
       "      <td>20709.783819</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2014-04-18</th>\n",
       "      <td>509.0000</td>\n",
       "      <td>474.25</td>\n",
       "      <td>483.77000</td>\n",
       "      <td>482.7500</td>\n",
       "      <td>482.7500</td>\n",
       "      <td>484.79</td>\n",
       "      <td>10458.045243</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2014-04-19</th>\n",
       "      <td>513.9899</td>\n",
       "      <td>473.83</td>\n",
       "      <td>505.01065</td>\n",
       "      <td>507.4999</td>\n",
       "      <td>502.5313</td>\n",
       "      <td>507.49</td>\n",
       "      <td>8963.618369</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                High     Low        Mid      Last       Bid     Ask  \\\n",
       "Date                                                                  \n",
       "2014-04-15  513.9000  452.00  504.23500  505.0000  503.5000  504.97   \n",
       "2014-04-16  547.0000  495.00  537.50000  538.0000  537.0000  538.00   \n",
       "2014-04-17  538.5000  486.10  507.02000  508.0000  506.0400  508.00   \n",
       "2014-04-18  509.0000  474.25  483.77000  482.7500  482.7500  484.79   \n",
       "2014-04-19  513.9899  473.83  505.01065  507.4999  502.5313  507.49   \n",
       "\n",
       "                  Volume  \n",
       "Date                      \n",
       "2014-04-15  21013.584774  \n",
       "2014-04-16  29633.358705  \n",
       "2014-04-17  20709.783819  \n",
       "2014-04-18  10458.045243  \n",
       "2014-04-19   8963.618369  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "data shape before embedding: (1445, 7)\n",
      "data shape after embedding: (1430, 15, 7)\n"
     ]
    }
   ],
   "source": [
    "print(\"data shape before embedding:\",data.shape)\n",
    "\n",
    "mean = data.mean(axis=0)\n",
    "std = data.std(axis=0)\n",
    "\n",
    "# zscore normalization\n",
    "data = ( data - mean ) / std\n",
    "\n",
    "# # of days past we want skynet to see\n",
    "d = 15\n",
    "\n",
    "X = np.zeros((data.shape[0],d,data.shape[1]))\n",
    "\n",
    "# embedding d days in each DP (deslocamento)\n",
    "for i in range(d,data.shape[0]):\n",
    "    X[i,:,:] = data.iloc[i-d:i].values\n",
    "\n",
    "#removing first d lines, this ones didn't have d days past\n",
    "X = X[d:,:,:]\n",
    "\n",
    "print(\"data shape after embedding:\",X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1445,)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# creating labels\n",
    "# in this day, was the coin worth more than 7 days ago ?\n",
    "Y = data['Mid'] - data.shift(7)['Mid']\n",
    "\n",
    "Y = Y > 0\n",
    "\n",
    "Y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#removing first line: second label refers to first DP ($$ delta)\n",
    "Y = Y[1:]\n",
    "\n",
    "#removing first d days because of the embedding\n",
    "Y = Y[d:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "#removing last registry from X, since we had to remove first DP from Y\n",
    "#specifying other dimensions for good practices - TY @lucasosouza\n",
    "X = X[: -1,:,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1429,), (1429, 15, 7))"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# checking shapes\n",
    "Y.shape, X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1429, 15, 7, 1)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Adding channel layer, as expected by the convnet\n",
    "X = X.reshape((Y.shape[0],d,7,1))\n",
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train / test split\n",
    "#X_train = X[:-300,:,:,:]\n",
    "#X_test = X[-300:,:,:,:]\n",
    "#Y_train = Y[:-300]\n",
    "#Y_test = Y[-300:]\n",
    "\n",
    "# split dat data \n",
    "kf = KFold(n_splits=8,shuffle=True,random_state=0)\n",
    "for train_index, test_index in kf.split(X):\n",
    "    X_train, X_test = X[train_index], X[test_index]\n",
    "    Y_train, Y_test = Y[train_index], Y[test_index]\n",
    "    #print(\"TRAIN:\", train_index, \"TEST:\", test_index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model(input_shape):\n",
    "    \n",
    "    X_input = Input(input_shape)\n",
    "    \n",
    "    #padding same so we dont lose size\n",
    "    X = Conv2D(32,(3,3), strides=(1,1),name=\"conv0\", padding=\"same\",kernel_regularizer=l2(0.0001))(X_input)\n",
    "    X = BatchNormalization(axis=3,name='bn0')(X)\n",
    "    X = Activation('relu')(X)\n",
    "\n",
    "    #X = MaxPooling2D((2,2),name='max_pool0')(X)\n",
    "    \n",
    "    #Second conv\n",
    "    X = Conv2D(64,(2,2), strides=(1,1),name=\"conv1\", padding=\"same\",kernel_regularizer=l2(0.0001))(X)\n",
    "    X = BatchNormalization(axis=3,name='bn1')(X)\n",
    "    X = Activation('relu')(X)\n",
    "\n",
    "    #X = MaxPooling2D((2,2),name='max_pool1')(X)\n",
    "    \n",
    "    #Third conv\n",
    "    X = Conv2D(128,(1,1), strides=(1,1),name=\"conv2\", padding=\"same\",kernel_regularizer=l2(0.0001))(X)\n",
    "    X = BatchNormalization(axis=3,name='bn2')(X)\n",
    "    X = Activation('relu')(X)\n",
    "    \n",
    "    #fcs\n",
    "    X = Flatten()(X)\n",
    "    X = Dense(128,activation=\"relu\")(X)\n",
    "    #X = Dropout(rate=0.3, seed=0)(X)\n",
    "    X = Dense(64,activation=\"relu\")(X)\n",
    "    #X = Dropout(rate=0.3, seed=0)(X)\n",
    "    X = Dense(32,activation=\"relu\")(X)\n",
    "    #X = Dropout(rate=0.3, seed=0)(X)\n",
    "    X = Dense(1,activation=\"sigmoid\")(X)\n",
    "    \n",
    "    model = Model(inputs=X_input,outputs=X, name=\"model1\")\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "1251/1251 [==============================] - 8s 6ms/step - loss: 0.9244 - acc: 0.5156\n",
      "Epoch 2/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.6877 - acc: 0.5787\n",
      "Epoch 3/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.6911 - acc: 0.5955\n",
      "Epoch 4/20\n",
      "1251/1251 [==============================] - 7s 5ms/step - loss: 0.6500 - acc: 0.6059\n",
      "Epoch 5/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.6586 - acc: 0.5971\n",
      "Epoch 6/20\n",
      "1251/1251 [==============================] - 7s 5ms/step - loss: 0.6500 - acc: 0.6187\n",
      "Epoch 7/20\n",
      "1251/1251 [==============================] - 7s 5ms/step - loss: 0.6514 - acc: 0.6083\n",
      "Epoch 8/20\n",
      "1251/1251 [==============================] - 7s 5ms/step - loss: 0.6200 - acc: 0.6355\n",
      "Epoch 9/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.6199 - acc: 0.6395\n",
      "Epoch 10/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.5909 - acc: 0.6587\n",
      "Epoch 11/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.5913 - acc: 0.6683\n",
      "Epoch 12/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.5630 - acc: 0.7034\n",
      "Epoch 13/20\n",
      "1251/1251 [==============================] - 7s 5ms/step - loss: 0.5397 - acc: 0.7114\n",
      "Epoch 14/20\n",
      "1251/1251 [==============================] - 9s 7ms/step - loss: 0.5178 - acc: 0.7226\n",
      "Epoch 15/20\n",
      "1251/1251 [==============================] - 8s 7ms/step - loss: 0.4973 - acc: 0.7490\n",
      "Epoch 16/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.4896 - acc: 0.7610\n",
      "Epoch 17/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.4771 - acc: 0.7634\n",
      "Epoch 18/20\n",
      "1251/1251 [==============================] - 7s 6ms/step - loss: 0.4470 - acc: 0.7898\n",
      "Epoch 19/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.4321 - acc: 0.7978\n",
      "Epoch 20/20\n",
      "1251/1251 [==============================] - 6s 5ms/step - loss: 0.3990 - acc: 0.8090\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f2b08124438>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#training the model, time for witchcraft\n",
    "model = model(X[0].shape)\n",
    "\n",
    "model.compile(optimizer='adam',loss='binary_crossentropy',metrics=[\"accuracy\"])\n",
    "\n",
    "model.fit(x = X_train, y = Y_train, epochs = 20, batch_size = 8,verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "178/178 [==============================] - 0s 1ms/step\n",
      "\n",
      "Loss = 0.559581062767\n",
      "Acc = 0.719101124265\n"
     ]
    }
   ],
   "source": [
    "preds = model.evaluate(x = X_test, y = Y_test)\n",
    "print()\n",
    "print(\"Loss = \" + str(preds[0]))\n",
    "print(\"Acc = \" + str(preds[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f2af4f31320>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f2b11a609e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data['Mid'].plot()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}