TIHM/TIHM_LSTM.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Example Pytorch Model Using the TIHM Pytorch Dataset:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# torch imports\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.utils import data as torchdata\n",
    "\n",
    "# sklearn imports\n",
    "from sklearn.utils import compute_sample_weight\n",
    "from sklearn import metrics\n",
    "\n",
    "# usual imports\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tqdm\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# loading dataset\n",
    "from data_loader import TIHMDataset\n",
    "\n",
    "DPATH = '../Dataset/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## load the data set\n",
    "train_dataset = TIHMDataset(\n",
    "    root=DPATH, train=True, normalise='global', n_days=7\n",
    "    )\n",
    "\n",
    "test_dataset = TIHMDataset(\n",
    "    root=DPATH, train=False, normalise='global', n_days=7\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The target names are:\n",
      "['Blood pressure', 'Agitation', 'Body water', 'Pulse', 'Weight', 'Body temperature']\n"
     ]
    }
   ],
   "source": [
    "print(f\"The target names are:\\n{train_dataset.target_names}\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This dataset can be passed to a dataloader: (we will apply some pre-processing and define these again later)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dl = torchdata.DataLoader(\n",
    "    dataset=train_dataset, batch_size=100, shuffle=True,\n",
    "    )\n",
    "\n",
    "test_dl = torchdata.DataLoader(\n",
    "    dataset=test_dataset, batch_size=100, shuffle=False,\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The total number of batches is 17\n"
     ]
    }
   ],
   "source": [
    "print(f\"The total number of batches is {len(train_dl)}\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us define the problem. We want to use the input data to predict there was agitation on a given day using the past 7 days worth of data. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "class AgitationDataset(torchdata.Dataset):\n",
    "    def __init__(self, dataset):\n",
    "        self.data, self.target = [], []\n",
    "        \n",
    "        for x, y in dataset:\n",
    "            self.data.append(torch.tensor(x).float())\n",
    "            self.target.append(np.int64(y[-1, 1]>=1)) # the last of the days and agitation\n",
    "        \n",
    "        # also define the sample weight\n",
    "        self.sw = compute_sample_weight(class_weight='balanced', y=self.target)\n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        return self.data[index], self.target[index], self.sw[index]\n",
    "    \n",
    "    def __len__(self):\n",
    "        return len(self.data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_agitation_dataset = AgitationDataset(train_dataset)\n",
    "test_agitation_dataset = AgitationDataset(test_dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dl = torchdata.DataLoader(\n",
    "    dataset=train_agitation_dataset, batch_size=100, shuffle=True,\n",
    "    )\n",
    "\n",
    "test_dl = torchdata.DataLoader(\n",
    "    dataset=test_agitation_dataset, batch_size=100, shuffle=False,\n",
    "    )"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# simple LSTM\n",
    "class LSTMModel(nn.Module):\n",
    "    def __init__(self, input_dim, sequence_length, hidden_size):\n",
    "        super().__init__()\n",
    "        self.layer1 = nn.Sequential(\n",
    "            nn.Linear(input_dim, input_dim),\n",
    "            nn.BatchNorm1d(num_features=sequence_length),\n",
    "            nn.ReLU(),\n",
    "            )\n",
    "        self.lstm = nn.LSTM(\n",
    "            input_dim, hidden_size, batch_first=True,\n",
    "            )\n",
    "        self.layer2 = nn.Sequential(\n",
    "            nn.Linear(hidden_size, 2)\n",
    "            )\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = self.layer1(x)\n",
    "        x, _ = self.lstm(x)\n",
    "        x = x[:, -1, :] # get the value from the last of the sequence\n",
    "        x = self.layer2(x)\n",
    "        return x"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And now train:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(model, criterion, optimiser, train_loader, n_epochs):\n",
    "    \n",
    "    training_loss = []\n",
    "\n",
    "    device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
    "    model.to(device)\n",
    "\n",
    "    def one_batch(x, y, sw):\n",
    "        x, y, sw = x.to(device), y.to(device), sw.to(device)\n",
    "        criterion.zero_grad()\n",
    "        output = model(x)\n",
    "        loss = criterion(output, y,)\n",
    "        loss = (loss * sw / sw.sum()).sum()\n",
    "        loss.mean().backward()\n",
    "        optimiser.step()\n",
    "        return loss\n",
    "\n",
    "    def one_epoch(train_loader):\n",
    "        epoch_loss = []\n",
    "        for batch in train_loader:\n",
    "            loss = one_batch(*batch)\n",
    "            epoch_loss.append(loss.item())\n",
    "        return epoch_loss\n",
    "\n",
    "    for epoch in tqdm.tqdm(range(n_epochs), desc='Training'):\n",
    "        epoch_loss = one_epoch(train_loader)\n",
    "        training_loss.extend(epoch_loss)\n",
    "\n",
    "    model.to('cpu')\n",
    "\n",
    "    return training_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_x = next(iter(train_dl))[0]\n",
    "lstm = LSTMModel(batch_x.shape[2], batch_x.shape[1], 64)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Training: 100%|██████████| 50/50 [00:07<00:00,  6.93it/s]\n"
     ]
    }
   ],
   "source": [
    "training_loss = train(\n",
    "    model=lstm, \n",
    "    criterion=nn.CrossEntropyLoss(reduction='none'), \n",
    "    optimiser=torch.optim.Adam(lstm.parameters(), lr=0.001),\n",
    "    train_loader=train_dl,\n",
    "    n_epochs=50,\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x1828162fa60>]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(training_loss)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And predict:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(model, test_loader):\n",
    "    device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
    "    model.to(device)\n",
    "\n",
    "    predictions = []\n",
    "    targets = []\n",
    "\n",
    "    for x, y, sw in test_loader:\n",
    "        x, y, sw = x.to(device), y.to(device), sw.to(device)\n",
    "        outputs = model(x)\n",
    "        predictions.append(F.softmax(outputs, dim=1))\n",
    "        targets.append(y)\n",
    "    \n",
    "    return torch.cat(predictions).to('cpu'), torch.cat(targets).to('cpu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions, targets = predict(lstm, test_dl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The confusion matrix is:\n",
      "[[365  44]\n",
      " [  8   5]]\n"
     ]
    }
   ],
   "source": [
    "cfm = metrics.confusion_matrix(targets, predictions.max(dim=1).indices)\n",
    "print(f\"The confusion matrix is:\\n{cfm}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The recall was 38.46%\n"
     ]
    }
   ],
   "source": [
    "recall = metrics.recall_score(targets, predictions.max(dim=1).indices)\n",
    "print(f\"The recall was {recall*100:.2f}%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The sensitivity is: 38.46%\n"
     ]
    }
   ],
   "source": [
    "sensitivity = cfm[1,1]/cfm[1,:].sum()\n",
    "print(f\"The sensitivity is: {sensitivity*100:.2f}%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The specificity is: 89.24%\n"
     ]
    }
   ],
   "source": [
    "specificity = cfm[0,0]/cfm[0,:].sum()\n",
    "print(f\"The specificity is: {specificity*100:.2f}%\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tihm",
   "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": "3.9.0"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "8ffa9458eb812bc86d8ec9c6172a4488b6504c6d9033306e2462dfb6e563f266"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}