P300 Classification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "bf1f4bd3",
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "63715ef1",
   "metadata": {},
   "outputs": [],
   "source": [
    "%%capture\n",
    "!pip install mne"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "bf512a13",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from scipy import signal\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.linear_model import RidgeClassifierCV, Ridge\n",
    "from sklearn.metrics import classification_report\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from mne.decoding import Vectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "9211f2bf",
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('./data/epochs.csv', index_col=[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "ed5ed65a",
   "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>time</th>\n",
       "      <th>condition</th>\n",
       "      <th>epoch</th>\n",
       "      <th>Fp1</th>\n",
       "      <th>Fp2</th>\n",
       "      <th>F5</th>\n",
       "      <th>AFz</th>\n",
       "      <th>F6</th>\n",
       "      <th>T7</th>\n",
       "      <th>Cz</th>\n",
       "      <th>T8</th>\n",
       "      <th>P7</th>\n",
       "      <th>P3</th>\n",
       "      <th>Pz</th>\n",
       "      <th>P4</th>\n",
       "      <th>P8</th>\n",
       "      <th>O1</th>\n",
       "      <th>Oz</th>\n",
       "      <th>O2</th>\n",
       "      <th>STI 014</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.097656</td>\n",
       "      <td>Target</td>\n",
       "      <td>0</td>\n",
       "      <td>-1.965512</td>\n",
       "      <td>43.550689</td>\n",
       "      <td>-18.469894</td>\n",
       "      <td>7.615624</td>\n",
       "      <td>13.062358</td>\n",
       "      <td>2.950307</td>\n",
       "      <td>5.193080</td>\n",
       "      <td>38.892224</td>\n",
       "      <td>2.482709</td>\n",
       "      <td>2.366828</td>\n",
       "      <td>-10.425305</td>\n",
       "      <td>6.175924</td>\n",
       "      <td>28.612600</td>\n",
       "      <td>15.837989</td>\n",
       "      <td>-0.008550</td>\n",
       "      <td>-5.403414</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-0.087891</td>\n",
       "      <td>Target</td>\n",
       "      <td>0</td>\n",
       "      <td>-8.878524</td>\n",
       "      <td>-0.753702</td>\n",
       "      <td>-39.762111</td>\n",
       "      <td>5.581617</td>\n",
       "      <td>-17.066895</td>\n",
       "      <td>-9.037337</td>\n",
       "      <td>-23.830406</td>\n",
       "      <td>19.622402</td>\n",
       "      <td>-41.573694</td>\n",
       "      <td>-27.536086</td>\n",
       "      <td>-45.680985</td>\n",
       "      <td>-20.230455</td>\n",
       "      <td>-17.798841</td>\n",
       "      <td>-18.669673</td>\n",
       "      <td>-28.362084</td>\n",
       "      <td>-27.623759</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.078125</td>\n",
       "      <td>Target</td>\n",
       "      <td>0</td>\n",
       "      <td>-7.923300</td>\n",
       "      <td>22.357380</td>\n",
       "      <td>-14.613151</td>\n",
       "      <td>1.043238</td>\n",
       "      <td>3.730958</td>\n",
       "      <td>-22.740073</td>\n",
       "      <td>-0.526012</td>\n",
       "      <td>12.032473</td>\n",
       "      <td>3.816798</td>\n",
       "      <td>-0.776339</td>\n",
       "      <td>-13.769359</td>\n",
       "      <td>-15.783225</td>\n",
       "      <td>-12.258964</td>\n",
       "      <td>10.163884</td>\n",
       "      <td>-5.595760</td>\n",
       "      <td>-5.650848</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-0.068359</td>\n",
       "      <td>Target</td>\n",
       "      <td>0</td>\n",
       "      <td>10.888668</td>\n",
       "      <td>2.224371</td>\n",
       "      <td>-22.415468</td>\n",
       "      <td>22.707990</td>\n",
       "      <td>2.656191</td>\n",
       "      <td>-8.252214</td>\n",
       "      <td>-3.393788</td>\n",
       "      <td>24.320502</td>\n",
       "      <td>-23.151776</td>\n",
       "      <td>-10.029675</td>\n",
       "      <td>-28.945462</td>\n",
       "      <td>-0.673688</td>\n",
       "      <td>13.192939</td>\n",
       "      <td>1.618473</td>\n",
       "      <td>-8.307013</td>\n",
       "      <td>-7.111478</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-0.058594</td>\n",
       "      <td>Target</td>\n",
       "      <td>0</td>\n",
       "      <td>-9.551013</td>\n",
       "      <td>11.443248</td>\n",
       "      <td>-17.984111</td>\n",
       "      <td>-4.119327</td>\n",
       "      <td>5.810256</td>\n",
       "      <td>-0.089011</td>\n",
       "      <td>-6.248183</td>\n",
       "      <td>33.287151</td>\n",
       "      <td>-9.872043</td>\n",
       "      <td>-13.848481</td>\n",
       "      <td>-19.698242</td>\n",
       "      <td>-8.886211</td>\n",
       "      <td>-3.499024</td>\n",
       "      <td>-4.661148</td>\n",
       "      <td>-11.851558</td>\n",
       "      <td>-11.887896</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       time condition  epoch        Fp1        Fp2         F5        AFz  \\\n",
       "0 -0.097656    Target      0  -1.965512  43.550689 -18.469894   7.615624   \n",
       "1 -0.087891    Target      0  -8.878524  -0.753702 -39.762111   5.581617   \n",
       "2 -0.078125    Target      0  -7.923300  22.357380 -14.613151   1.043238   \n",
       "3 -0.068359    Target      0  10.888668   2.224371 -22.415468  22.707990   \n",
       "4 -0.058594    Target      0  -9.551013  11.443248 -17.984111  -4.119327   \n",
       "\n",
       "          F6         T7         Cz         T8         P7         P3  \\\n",
       "0  13.062358   2.950307   5.193080  38.892224   2.482709   2.366828   \n",
       "1 -17.066895  -9.037337 -23.830406  19.622402 -41.573694 -27.536086   \n",
       "2   3.730958 -22.740073  -0.526012  12.032473   3.816798  -0.776339   \n",
       "3   2.656191  -8.252214  -3.393788  24.320502 -23.151776 -10.029675   \n",
       "4   5.810256  -0.089011  -6.248183  33.287151  -9.872043 -13.848481   \n",
       "\n",
       "          Pz         P4         P8         O1         Oz         O2  STI 014  \n",
       "0 -10.425305   6.175924  28.612600  15.837989  -0.008550  -5.403414      0.0  \n",
       "1 -45.680985 -20.230455 -17.798841 -18.669673 -28.362084 -27.623759      0.0  \n",
       "2 -13.769359 -15.783225 -12.258964  10.163884  -5.595760  -5.650848      0.0  \n",
       "3 -28.945462  -0.673688  13.192939   1.618473  -8.307013  -7.111478      0.0  \n",
       "4 -19.698242  -8.886211  -3.499024  -4.661148 -11.851558 -11.887896      0.0  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "9c6e1273",
   "metadata": {},
   "outputs": [],
   "source": [
    "def split_by_channels(data):\n",
    "    columns = data.columns\n",
    "    result = []\n",
    "\n",
    "    for col in columns:\n",
    "        result.append(data[col].to_list())\n",
    "\n",
    "    return np.array(result, np.float)\n",
    "\n",
    "def to_record(data, epoch_idx):\n",
    "    epoch_data = data[(data['epoch'] == i) & (data['STI 014'] <= 0)]\n",
    "    exclude_cols = ~data.columns.isin(['time', 'condition', 'epoch', 'STI 014'])\n",
    "\n",
    "    label = 0. if epoch_data['condition'].iloc[-1] == 'NonTarget' else 1.\n",
    "    values = split_by_channels(epoch_data.loc[:, exclude_cols])\n",
    "\n",
    "    return (values, label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "1d968088",
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset_X = None\n",
    "dataset_y = None\n",
    "\n",
    "if os.path.isfile('./data/dataset.npz'):\n",
    "    with np.load('./data/dataset.npz') as file:\n",
    "        dataset_X, dataset_y = file['dataset_X'], file['dataset_y']\n",
    "else:\n",
    "    for i in data['epoch'].unique():\n",
    "        X, y = to_record(data, i)\n",
    "        if dataset_X is None:\n",
    "            dataset_X = np.array([X[:, :99]], np.float)\n",
    "        else:\n",
    "            dataset_X = np.append(dataset_X, [X[:, :99]], axis=0)\n",
    "\n",
    "        if dataset_y is None:\n",
    "            dataset_y = np.array([y], np.float)\n",
    "        else:\n",
    "            dataset_y = np.append(dataset_y, [y], axis=0)\n",
    "            \n",
    "    np.savez_compressed('./data/dataset.npz', dataset_X=dataset_X, dataset_y=dataset_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "76cee020",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_X, test_X, train_y, test_y = train_test_split(dataset_X, dataset_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "f0a4c01f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.base import BaseEstimator, TransformerMixin\n",
    "\n",
    "class Transformer(BaseEstimator, TransformerMixin):\n",
    "    '''\n",
    "    Base class for transformers providing dummy implementation\n",
    "        of the methods expected by sklearn\n",
    "    '''\n",
    "    def fit(self, x, y=None):\n",
    "        return self\n",
    "\n",
    "class ButterFilter(Transformer):\n",
    "    '''Applies Scipy's Butterworth filter'''\n",
    "    def __init__(self, sampling_rate: int, order: int, highpass: int, lowpass: int) -> None:\n",
    "        self.sampling_rate = sampling_rate\n",
    "        self.order = order\n",
    "        self.highpass = highpass\n",
    "        self.lowpass = lowpass\n",
    "\n",
    "        normal_cutoff = [a / (0.5 * self.sampling_rate) for a in (self.highpass, self.lowpass)]\n",
    "        self.filter = signal.butter(self.order, normal_cutoff, btype='bandpass')\n",
    "\n",
    "    def transform(self, x):\n",
    "        out = np.empty_like(x)\n",
    "        out[:] = [signal.filtfilt(*self.filter, item) for item in x]\n",
    "        return out\n",
    "\n",
    "class ChannellwiseScaler(Transformer):\n",
    "    '''Performs channelwise scaling according to given scaler\n",
    "    '''\n",
    "    def __init__(self, scaler: Transformer):\n",
    "        '''Args:\n",
    "            scaler: instance of one of sklearn.preprocessing classes\n",
    "                StandardScaler or MinMaxScaler or analogue\n",
    "        '''\n",
    "        self.scaler = scaler\n",
    "\n",
    "    def fit(self, x: np.ndarray, y=None):\n",
    "        '''\n",
    "        Args:\n",
    "            x: array of eegs, that is every element of x is (n_channels, n_ticks)\n",
    "                x shaped (n_eegs) of 2d array or (n_eegs, n_channels, n_ticks)\n",
    "        '''\n",
    "        for signals in x:\n",
    "            self.scaler.partial_fit(signals.T)\n",
    "        return self\n",
    "\n",
    "    def transform(self, x):\n",
    "        '''Scales each channel\n",
    "\n",
    "        Wors either with one record, 2-dim input, (n_channels, n_samples)\n",
    "            or many records 3-dim, (n_records, n_channels, n_samples)\n",
    "        Returns the same format as input\n",
    "        '''\n",
    "        scaled = np.empty_like(x)\n",
    "        for i, signals in enumerate(x):\n",
    "            # double T for scaling each channel separately\n",
    "            scaled[i] = self.scaler.transform(signals.T).T\n",
    "        return scaled"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "694c1ffb",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Projects\\ml\\moabb-bi2013-reformat\\venv\\lib\\site-packages\\sklearn\\linear_model\\_logistic.py:444: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
      "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
      "\n",
      "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
      "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
      "Please also refer to the documentation for alternative solver options:\n",
      "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
      "  n_iter_i = _check_optimize_result(\n",
      "C:\\Projects\\ml\\moabb-bi2013-reformat\\venv\\lib\\site-packages\\sklearn\\linear_model\\_logistic.py:444: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
      "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
      "\n",
      "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
      "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
      "Please also refer to the documentation for alternative solver options:\n",
      "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
      "  n_iter_i = _check_optimize_result(\n",
      "C:\\Projects\\ml\\moabb-bi2013-reformat\\venv\\lib\\site-packages\\sklearn\\linear_model\\_logistic.py:444: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
      "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
      "\n",
      "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
      "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
      "Please also refer to the documentation for alternative solver options:\n",
      "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
      "  n_iter_i = _check_optimize_result(\n",
      "C:\\Projects\\ml\\moabb-bi2013-reformat\\venv\\lib\\site-packages\\sklearn\\linear_model\\_logistic.py:444: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
      "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
      "\n",
      "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
      "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
      "Please also refer to the documentation for alternative solver options:\n",
      "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
      "  n_iter_i = _check_optimize_result(\n",
      "C:\\Projects\\ml\\moabb-bi2013-reformat\\venv\\lib\\site-packages\\sklearn\\linear_model\\_logistic.py:444: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
      "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
      "\n",
      "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
      "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
      "Please also refer to the documentation for alternative solver options:\n",
      "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
      "  n_iter_i = _check_optimize_result(\n",
      "C:\\Projects\\ml\\moabb-bi2013-reformat\\venv\\lib\\site-packages\\sklearn\\linear_model\\_logistic.py:444: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
      "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
      "\n",
      "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
      "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
      "Please also refer to the documentation for alternative solver options:\n",
      "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
      "  n_iter_i = _check_optimize_result(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.92      0.96      0.94      3359\n",
      "         1.0       0.70      0.51      0.59       599\n",
      "\n",
      "    accuracy                           0.89      3958\n",
      "   macro avg       0.81      0.74      0.76      3958\n",
      "weighted avg       0.88      0.89      0.89      3958\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis\n",
    "from sklearn.linear_model import RidgeClassifierCV\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.svm import LinearSVC, SVC\n",
    "from sklearn.linear_model import LogisticRegressionCV\n",
    "\n",
    "preproc = make_pipeline(\n",
    "    ButterFilter(512 // 5, 4, 0.5, 20), \n",
    "    ChannellwiseScaler(StandardScaler()),\n",
    "    Vectorizer(),\n",
    ")\n",
    "\n",
    "preproc.fit(train_X)\n",
    "\n",
    "# cls = MLPClassifier(hidden_layer_sizes=(1000, 500), learning_rate='adaptive', \n",
    "#                    activation='relu')\n",
    "# cls = LinearDiscriminantAnalysis(solver='eigen', shrinkage='auto')\n",
    "# cls = RidgeClassifierCV()\n",
    "# cls = LinearSVC(C=4)\n",
    "cls = LogisticRegressionCV()\n",
    "cls.fit(preproc.transform(train_X), train_y)\n",
    "\n",
    "preds = cls.predict(preproc.transform(test_X))\n",
    "print(classification_report(test_y, preds))"
   ]
  }
 ],
 "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": "3.10.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}