Forecasting_Demo.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bcc35afe475747b8adac1809fbcf9bf0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Dropdown(description='Utility/Building', options=(('CHAPEL_15MIN_STEAM', 'CHAPEL_15MIN_STEAM'), ('CHAPEL_15MIN…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1e57585375be4042afcb138ddba721bb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=False, description='Clean Data with Anomaly Detection', disabled=True, layout=Layout(object_pos…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "00fbf4af5fe640019b19286214e6923f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Dropdown(description='Aggregation Level', disabled=True, options=(('Monthly', 'month'), ('Weekly', 'week'), ('…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "677d4079413546398c70d2715d4fb31c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "BoundedIntText(value=12, description='Forecast Length (Months):', max=60, min=1, style=DescriptionStyle(descri…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b498dec5a4ad47e9b2547e7cd4ab4526",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(Button(button_style='success', description='Display Forecast', layout=Layout(display='inline-fl…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "de0b3e8263ad46e681f6359a7e528693",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Output()"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.colors as colors\n",
    "import ipywidgets as ipw\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import numpy.ma as ma\n",
    "import sesd\n",
    "from datetime import datetime\n",
    "from tqdm import tqdm\n",
    "from IPython.display import clear_output\n",
    "from halo import HaloNotebook as Halo # For spinner\n",
    "import time\n",
    "\n",
    "from pandas.plotting import autocorrelation_plot\n",
    "from statsmodels.tsa.arima.model import ARIMA\n",
    "from statsmodels.tsa.stattools import adfuller\n",
    "from statsmodels.graphics.tsaplots import plot_pacf\n",
    "from sklearn.linear_model import LinearRegression\n",
    "\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\") #UI can get fucked if any warnings come up.\n",
    "\n",
    "# To prevent automatic figure display when execution of the cell ends\n",
    "#%matplotlib notebook\n",
    "\n",
    "\n",
    "# In[2]:\n",
    "\n",
    "\n",
    "# Read data and some basic processing\n",
    "# this will be replaced by a file picker and some library code in the non-demo version\n",
    "data = pd.read_csv(\"https://raw.githubusercontent.com/epswartz/sample_forecasting_tool/master/sample_data.csv\")\n",
    "fmt = \"%m/%d/%y %I:%M %p\"\n",
    "data.time = pd.to_datetime(data.time, format=fmt)\n",
    "\n",
    "# Aggregate all 3 into monthly sums\n",
    "year_month_start = 1 + (2017*12)\n",
    "data['monthly-time'] = data.time.apply(lambda x: (x.year*12) + x.month - year_month_start)\n",
    "\n",
    "util_series_names = data.drop(['time', 'monthly-time'], axis=1).columns\n",
    "\n",
    "\n",
    "# In[3]:\n",
    "\n",
    "\n",
    "# Declare all the widgets\n",
    "util_picker = ipw.Dropdown(\n",
    "    style={'description_width': 'initial'},\n",
    "    options=[\n",
    "        ('Steam', 'stm'),\n",
    "        ('Chilled Water', 'chw'),\n",
    "        ('Hot Water', 'hhw')\n",
    "    ],\n",
    "    value='stm',\n",
    "    description=\"Utility Type:\"\n",
    ")\n",
    "\n",
    "# Initially blank because the options are read in from file\n",
    "building_picker = ipw.Dropdown(\n",
    "    style={'description_width': 'initial'},\n",
    "    options=[(n,n) for n in util_series_names],\n",
    "    description=\"Utility/Building\"\n",
    "   #description=\"Building/Collection Point:\"\n",
    ")\n",
    "\n",
    "anom_toggle = ipw.Checkbox(\n",
    "    style={'description_width': 'initial'},\n",
    "    value=False,\n",
    "    description='Clean Data with Anomaly Detection',\n",
    "    disabled=True,\n",
    "    layout={'object_position':'left'}\n",
    ")\n",
    "\n",
    "aggregation_picker = ipw.Dropdown(\n",
    "    style={'description_width': 'initial'},\n",
    "    options=[\n",
    "        ('Monthly', 'month'),\n",
    "        ('Weekly', 'week'),\n",
    "        ('Daily', 'day')\n",
    "    ],\n",
    "    value='month',\n",
    "    description=\"Aggregation Level\",\n",
    "    disabled=True # For demo version only monthly\n",
    ")\n",
    "\n",
    "forecast_len_field = ipw.BoundedIntText(\n",
    "    style={'description_width': 'initial'},\n",
    "    value=12,\n",
    "    min=1,\n",
    "    max=60,\n",
    "    step=1,\n",
    "    description='Forecast Length (Months):',\n",
    "    disabled=False\n",
    ")\n",
    "\n",
    "btn_layout = ipw.Layout(width='auto', height='40px', display=\"inline-flex\")\n",
    "\n",
    "display_btn = ipw.Button(\n",
    "    layout=btn_layout,\n",
    "    style={'button_color': 'lightgreen'},\n",
    "    description=\"Display Forecast\",\n",
    "    disabled=False,\n",
    "    button_style='success', # 'success', 'info', 'warning', 'danger' or ''\n",
    "    tooltip='After choosing settings, click to display forecast time chart',\n",
    "    #icon='fa-line-chart'\n",
    ")\n",
    "\n",
    "export_btn = ipw.Button(\n",
    "    layout=btn_layout,\n",
    "    style={'button_color': 'lightskyblue'},\n",
    "    description=\"Export Forecast CSV\",\n",
    "    disabled=True,\n",
    "    button_style='success', # 'success', 'info', 'warning', 'danger' or ''\n",
    "    tooltip=\"After choosing settings, click to export data\",\n",
    "    #icon='fa-save'\n",
    ")\n",
    "\n",
    "\n",
    "clear_btn = ipw.Button(\n",
    "    layout=btn_layout,\n",
    "    style={'button_color': '#f05c79'},\n",
    "    description=\"Clear Plots\",\n",
    "    disabled=False,\n",
    "    button_style='success', # 'success', 'info', 'warning', 'danger' or ''\n",
    "    tooltip=\"Deletes all currently shown plots.\"\n",
    "    #icon='fa-save'\n",
    ")\n",
    "\n",
    "btn_box = ipw.HBox([display_btn, export_btn, clear_btn])\n",
    "\n",
    "figure_display_area = ipw.Output()\n",
    "\n",
    "\n",
    "# In[4]:\n",
    "\n",
    "\n",
    "# Graphics/UI and direct handler functions\n",
    "def render():\n",
    "    \"\"\"\n",
    "    Becuase the content of some UI elements depends on the settings in others,\n",
    "    the UI needs to be re-displayed whenever something is changed.\n",
    "    \n",
    "    This gets called once at program start, and then again after every setting change.\n",
    "    \"\"\"\n",
    "    \n",
    "    #display(util_picker) # No util picker for demo.\n",
    "    \n",
    "    # Get building options from loaded file\n",
    "    building_picker.options = [(n,n) for n in util_series_names]\n",
    "    display(building_picker)\n",
    "    \n",
    "    display(anom_toggle)\n",
    "    \n",
    "    display(aggregation_picker)\n",
    "    \n",
    "    display(forecast_len_field)\n",
    "    \n",
    "    # TODO build and display normalization options for external data\n",
    "    \n",
    "    display(btn_box)\n",
    "    \n",
    "    display(figure_display_area)\n",
    "\n",
    "def on_change():\n",
    "    \"\"\"\n",
    "    Called whenever any UI element is changed.\n",
    "    \n",
    "    Responsible for changing active_series, an identifier (file and column name) for the series of interest.\n",
    "    Also responsible for calling render().\n",
    "    \"\"\"\n",
    "    render()\n",
    "    \n",
    "# TODO split the data processing part of this out into something that gets called by this and\n",
    "# the export handler.\n",
    "def display_forecast_on_click(b):\n",
    "    \"\"\"\n",
    "    Runs whenever the \"Display Forecast\" button is clicked.\n",
    "    \n",
    "    Does the forecasting if needed, and show the plot.\n",
    "    \"\"\"\n",
    "    global current_forecast_key\n",
    "    global current_forecast_ts\n",
    "    global current_aggregated_data\n",
    "    \n",
    "    \n",
    "    #spinner = Halo(text='Loading', spinner='dots')\n",
    "    #spinner.start()\n",
    "    \n",
    "    # If we've already done the calculation, just do the plotting.\n",
    "    # FIXME This needs to check other settings\n",
    "    if (current_forecast_key == building_picker.value) and (forecast_len_field.value == len(current_forecast_ts)):\n",
    "        show_forecast_plot()\n",
    "        return\n",
    "    \n",
    "    current_forecast_key = building_picker.value\n",
    "    \n",
    "    # Aggregate data into monthly and convert to numpy\n",
    "    monthly_sum = data.groupby(['monthly-time'])[building_picker.value].agg('sum')\n",
    "    ts = monthly_sum.to_numpy()\n",
    "    current_aggregated_data = ts\n",
    "    \n",
    "    # Make forecast with ARIMA\n",
    "    model = ARIMA(ts, order=(12,1,0))\n",
    "    model_fit = model.fit()\n",
    "    \n",
    "    current_forecast_ts = model_fit.forecast(steps=forecast_len_field.value)\n",
    "    \n",
    "    show_forecast_plot()\n",
    "    \n",
    "    #spinner.stop()\n",
    "\n",
    "\n",
    "# Have to wrap the clear fn to take care of the param\n",
    "def clear_btn_on_click(b):\n",
    "    plt.close('all')\n",
    "    figure_display_area.clear_output()\n",
    "\n",
    "\n",
    "# In[5]:\n",
    "\n",
    "\n",
    "# Functions for plotting\n",
    "\n",
    "def show_forecast_plot():\n",
    "    with figure_display_area:\n",
    "        #plt.close('all')\n",
    "        #figure_display_area.clear_output() # Clear the old plot\n",
    "        \n",
    "        # Something north of here is async and I don't seem to have any API to mess with it,\n",
    "        # but it makes the UI fuck up sometimes. For now, I'm going to add a 1 second sleep\n",
    "        # and just pray that nobody comes to (rightfully) kill me in my sleep.\n",
    "        #time.sleep(1) # God help me.\n",
    "        \n",
    "        plt.figure(figsize=(12,8))\n",
    "        plt.title(f\"Forecasting: {current_forecast_key}\", size=15)\n",
    "        plt.plot(current_aggregated_data, label=\"Training Data\", marker=\"o\")\n",
    "        plt.plot(np.arange(forecast_len_field.value) + len(current_aggregated_data), current_forecast_ts, color=\"orange\", label=\"Forecast\", marker=\"o\")\n",
    "        plt.ylabel(\"Utility Usage\", size=12)\n",
    "        plt.legend()\n",
    "        plt.show()\n",
    "\n",
    "\n",
    "# In[6]:\n",
    "\n",
    "\n",
    "# Functions for data cleaning and imputation\n",
    "# It's possible this should be a class inside a module - let's discuss that with Grace\n",
    "\n",
    "def clean_data():\n",
    "    pass\n",
    "\n",
    "\n",
    "# In[7]:\n",
    "\n",
    "\n",
    "# Functions for forecasting\n",
    "# This should almost certainly be a module - discuss with Shota\n",
    "\n",
    "def forecast_sarimax():\n",
    "    pass\n",
    "\n",
    "\n",
    "# In[8]:\n",
    "\n",
    "\n",
    "# Configure on_click and on_change behaviors\n",
    "display_btn.on_click(display_forecast_on_click)\n",
    "clear_btn.on_click(clear_btn_on_click)\n",
    "\n",
    "# Dummy forecast globals\n",
    "current_forecast_key = \"\"\n",
    "current_forecast_ts = np.asarray([])\n",
    "current_aggregated_data = np.asarray([])\n",
    "\n",
    "# Start the app\n",
    "render()\n",
    "\n",
    "# May need to somehow declare current_forecast_ts_df and current_forecast_key here.\n",
    "\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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}