{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Notebook: Forecast Visualization\n",
    "\n",
    "This notebook gives an example visualization of a demand dataset and related timeseries data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set the start and end dates of the forecast\n",
    "start_date = '2014-09-01' # YYYY-MM-DD\n",
    "end_date = '2015-02-01'   # YYYY-MM-DD\n",
    "\n",
    "# provide the full CSV name uploaded to the /train folder in S3\n",
    "demand_dataset_name = 'DATASET_NAME.csv'\n",
    "related_dataset_name = 'DATASET_NAME.related.csv'\n",
    "\n",
    "# provide the exports to show from the export/ folder in S3 (these are created by Amazon Forecast)\n",
    "forecast_exports = [\n",
    "    {\n",
    "        'path': 'export_date/export_name.csv',\n",
    "        'name': 'Forecast 1 - Name'\n",
    "    },    \n",
    "    {\n",
    "        'path': 'export_date/export_name.csv',\n",
    "        'name': 'Forecast 2 - Name'\n",
    "    }\n",
    "    \n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os \n",
    "import pandas as pd\n",
    "\n",
    "forecast_bucket = os.getenv('FORECAST_BUCKET')\n",
    "\n",
    "def download_data(path, date_column, header=None):\n",
    "    data = pd.read_csv(f\"s3://{forecast_bucket}/{path}\", header=header, parse_dates=[date_column])\n",
    "    \n",
    "    # filter the data to the dates specified \n",
    "    flt = (data[date_column] >= start_date) & (data[date_column] <= end_date)\n",
    "    data = data.loc[flt]\n",
    "    \n",
    "    return data\n",
    "\n",
    "def get_exports(exports_list, date_column):\n",
    "    for export in exports_list:\n",
    "        export['data'] = download_data(f\"exports/{export.get('path')}\", date_column, header=0)\n",
    "\n",
    "\n",
    "demand = download_data(f\"train/{demand_dataset_name}\", date_column=1)\n",
    "relate = download_data(f\"train/{related_dataset_name}\", date_column=1)\n",
    "get_exports(forecast_exports, date_column='date')  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.dates as mdates\n",
    "import matplotlib.pyplot as plt\n",
    "import re \n",
    "\n",
    "years = mdates.YearLocator()\n",
    "months = mdates.MonthLocator()\n",
    "years_fmt = mdates.DateFormatter(\"%Y\")\n",
    "\n",
    "\n",
    "# there will be a subplot for every export, and an extra for the related timeseries\n",
    "fig, axes = plt.subplots(len(forecast_exports) + 1, sharex=True)\n",
    "fig.set_size_inches(18.5, 10.5)\n",
    "\n",
    "axes[0].xaxis.set_major_locator(years)\n",
    "axes[0].xaxis.set_major_formatter(years_fmt)\n",
    "axes[0].xaxis.set_minor_locator(months)\n",
    "axes[0].format_xdata = mdates.DateFormatter(\"%Y-%m-%d\")\n",
    "\n",
    "for idx, ax in enumerate(axes): \n",
    "    ax.grid(True)\n",
    "    \n",
    "    if idx >= len(forecast_exports):\n",
    "        continue\n",
    "    \n",
    "    title = forecast_exports[idx].get('name')\n",
    "    data = forecast_exports[idx].get('data')\n",
    "    \n",
    "    ax.set_title(title)\n",
    "    ax.set_xlabel(\"Date\")\n",
    "    ax.set_ylabel(\"Demand\")\n",
    "    \n",
    "    # plot the demand\n",
    "    ax.plot(demand[1], demand[2], linestyle='solid', color='DodgerBlue', label=\"Demand\")\n",
    "    \n",
    "    # plot a dashed line from the end of the data to the start of the forecast\n",
    "    xs = [demand[1].iloc[-1], data.date.iloc[0]]\n",
    "    ys = [demand[2].iloc[-1], data.p50.iloc[0]]\n",
    "    ax.plot(xs, ys, linestyle='dashed', color='DodgerBlue')\n",
    "\n",
    "    # plot each forecast\n",
    "    ax.plot(data.date, data.p50, linestyle='dashed', color='DodgerBlue', label='P50')    \n",
    "\n",
    "    # plot the quantiles\n",
    "    colors = ['LightBlue', 'LightSteelBlue']\n",
    "    for pnn in [pnn for pnn in data.columns if re.match('^p\\d+$', pnn) and pnn != 'p50']:\n",
    "        color = colors.pop()\n",
    "        colors.insert(0, color)\n",
    "        \n",
    "        ax.fill_between(data.date, data.p50, data[pnn], color=color, label=pnn)\n",
    "\n",
    "    ax.legend(loc=\"lower left\")\n",
    "\n",
    "    \n",
    "# this will show a related timeseries on the same plot using the same x axis \n",
    "axes[-1].set_title(\"Item Price\")\n",
    "axes[-1].set_xlabel(\"Date\")\n",
    "axes[-1].set_ylabel(\"Price ($)\")\n",
    "axes[-1].plot(relate[1], relate[2])\n",
    "\n",
    "fig.autofmt_xdate()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "conda_python3",
   "language": "python",
   "name": "conda_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": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}