notebooks/delphi_example_demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# DELPHI: Example Demonstration"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load dependency packages\n",
    "Note that all of the necessary dependency packages should have been automatically installed if you are running this notebook from within the provided Docker container. If not, you can install the dependencies using the provided `requirements.txt` file using pip: \n",
    "```\n",
    "pip install --upgrade pip && pip install -U -r requirements.txt\n",
    "```\n",
    "This notebook assumes familiarity with [Pandas](https://pandas.pydata.org/docs/) and [Seaborn](https://seaborn.pydata.org). In addition, knowledge of [scikit-learn](https://scikit-learn.org/stable/user_guide.html) may be helpful. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 349,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import seaborn as sns\n",
    "\n",
    "sns.set(style='ticks')\n",
    "\n",
    "sns.set_palette(\"tab10\")\n",
    "\n",
    "sns.set_context('paper')\n",
    "\n",
    "from sklearn.metrics import balanced_accuracy_score, roc_auc_score, confusion_matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load demonstration models and data\n",
    "Binary files with models and data are available for download (links in `README.md`). These files should be downloaded and added to the Docker container where they can be accessed below. Alternatively, you can provide your own data here. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 350,
   "metadata": {},
   "outputs": [],
   "source": [
    "fname = '/tmp/data/demo_models.pickle'    \n",
    "\n",
    "with open(fname, 'rb') as f:\n",
    "    \n",
    "    mdls = pickle.load(f)\n",
    "    \n",
    "fname = '/tmp/data/demo_data.pickle'         \n",
    "\n",
    "with open(fname, 'rb') as f:\n",
    "    \n",
    "    d = pickle.load(f)    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run demonstration predictions on example data "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Example models and data, along with true values, are provided for this analysis. The example uses data from two-years post-publication, and explores the early-warning signal calculation for the following three years using a subset of the data and features. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 351,
   "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>lo</th>\n",
       "      <th>hi</th>\n",
       "      <th>y1</th>\n",
       "      <th>y2</th>\n",
       "      <th>p</th>\n",
       "      <th>t</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.088477</td>\n",
       "      <td>0.911523</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.558275</td>\n",
       "      <td>0.441725</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>False</td>\n",
       "      <td>False</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>295</th>\n",
       "      <td>0.470591</td>\n",
       "      <td>0.529409</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>296</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>297</th>\n",
       "      <td>0.816461</td>\n",
       "      <td>0.183539</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>298</th>\n",
       "      <td>0.934911</td>\n",
       "      <td>0.065089</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>299</th>\n",
       "      <td>0.355556</td>\n",
       "      <td>0.644444</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>300 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           lo        hi  y1  y2      p      t\n",
       "0    1.000000  0.000000   2   3  False  False\n",
       "1    0.088477  0.911523   2   3   True   True\n",
       "2    0.558275  0.441725   2   3  False  False\n",
       "3    1.000000  0.000000   2   3  False  False\n",
       "4    1.000000  0.000000   2   3  False  False\n",
       "..        ...       ...  ..  ..    ...    ...\n",
       "295  0.470591  0.529409   2   5   True   True\n",
       "296  1.000000  0.000000   2   5  False  False\n",
       "297  0.816461  0.183539   2   5  False  False\n",
       "298  0.934911  0.065089   2   5  False  False\n",
       "299  0.355556  0.644444   2   5  False   True\n",
       "\n",
       "[300 rows x 6 columns]"
      ]
     },
     "execution_count": 351,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame()\n",
    "mdf = pd.DataFrame()\n",
    "\n",
    "for ((y1, y2), m), (X, y) in zip(mdls.items(), d.values()):\n",
    "    \n",
    "    # Calculate predictions for example data points and save\n",
    "    p = m.predict_proba(X)\n",
    "    \n",
    "    t = m.predict(X)\n",
    "    \n",
    "    idf = pd.DataFrame(p, columns=['lo','hi'])\n",
    "    \n",
    "    idf['y1'] = y1\n",
    "    \n",
    "    idf['y2'] = y2\n",
    "\n",
    "    idf['p'] = y\n",
    "    \n",
    "    idf['t'] = t\n",
    "    \n",
    "    df = pd.concat([df, idf], ignore_index=True)\n",
    "    \n",
    "    # Calculate evaluation metrics for example models and save\n",
    "    mdf = mdf.append([{\n",
    "        'ba': balanced_accuracy_score(y, t),\n",
    "        'roc': roc_auc_score(y, p[:,1]),\n",
    "        'y1': y1,\n",
    "        'y2': y2\n",
    "    }], ignore_index=True)\n",
    "    \n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 352,
   "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>ba</th>\n",
       "      <th>roc</th>\n",
       "      <th>y1</th>\n",
       "      <th>y2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.954082</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.943299</td>\n",
       "      <td>0.924399</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         ba       roc  y1  y2\n",
       "0  1.000000  1.000000   2   3\n",
       "1  0.954082  1.000000   2   4\n",
       "2  0.943299  0.924399   2   5"
      ]
     },
     "execution_count": 352,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mdf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Explore results\n",
    "Now that we have predictions on our example dataset, we can explore the performance of the provided example DELPHI models on our sample dataset. Note that because our example here is a small dataset, the variance is higher that it would be with a larger dataset. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluation metrics\n",
    "\n",
    "We can also explore the trends in predictive performance over time (i.e. as the years post-publication increases). As the dataset size increases, these values converge to the values demonstrated in the manuscript. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 371,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 979.68x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "mmdf = mdf.drop('y1', axis=1)\n",
    "\n",
    "mmdf = mmdf.rename(columns={'ba':'Balanced Accuracy', 'roc':'AUC ROC', 'y2':'Year Post-Publication'})\n",
    "\n",
    "mmdf = mmdf.melt(id_vars='Year Post-Publication', var_name='Evaluation Metric', value_name='Score')\n",
    "\n",
    "\n",
    "g = sns.catplot(data=mmdf, x='Year Post-Publication', y='Score', hue='Evaluation Metric', height=8, \n",
    "                kind='point', aspect=1.5, legend=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Confusion matrices"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can easily plot and visualize confusion matrices for the performance of our example models on our example data. For clarity, a confusion matrix $C$ has entries $C_{i,j}$ equal to the observations from group $i$ predicted to be in group $j$. In our case, both $i$ and $j$ can be either high-impact or low-impact. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 353,
   "metadata": {},
   "outputs": [],
   "source": [
    "cdf = pd.DataFrame()\n",
    "\n",
    "for i in df.y2.unique():\n",
    "    \n",
    "    idf = df[df.y2==i]\n",
    "    \n",
    "    cm = confusion_matrix(idf.p, idf.t)\n",
    "    \n",
    "    cdf = cdf.append(pd.Series(cm.ravel()), ignore_index=True)\n",
    "\n",
    "cdf.columns = ['True Negative', 'False Positive', 'False Negative', 'True Positive']\n",
    "\n",
    "cdf['Year Post-Publication'] = df.y2.unique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 354,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def draw_heatmap(*args, **kwargs):\n",
    "    \n",
    "    data = kwargs.pop('data')\n",
    "    \n",
    "    d = data.iloc[:,:4].values.reshape(2,2)\n",
    "    \n",
    "    sns.heatmap(d, **kwargs)\n",
    "    \n",
    "fg = sns.FacetGrid(data=cdf, col='Year Post-Publication', height=5)\n",
    "\n",
    "ls = ['Low-Impact', 'High-Impact']\n",
    "\n",
    "fg.map_dataframe(draw_heatmap, cbar=False, annot=True, cmap='OrRd', linewidth=1/2, \n",
    "                 square=True, xticklabels=ls, yticklabels=ls)\n",
    "\n",
    "_ = fg.set_axis_labels(\"Prediction\", \"Ground Truth\")"
   ]
  }
 ],
 "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}