Statsmodels - Dynamic Prediction

dynamic_pred.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import statsmodels.api as sm\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Get some data\n",
    "macrodata = sm.datasets.macrodata.load_pandas().data\n",
    "macrodata.index = pd.PeriodIndex(start='1959Q1', end='2009Q3', freq='Q')\n",
    "endog = macrodata['infl'] + 0.1\n",
    "\n",
    "# Estimate the parameters of the model\n",
    "mod = sm.tsa.UnobservedComponents(endog, 'llevel', seasonal=4)\n",
    "res = mod.fit()\n",
    "\n",
    "np.set_printoptions(suppress=True, precision=5, linewidth=120)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dynamic prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_forecasts = 10\n",
    "pred1 = np.zeros((n_forecasts, mod.nobs))\n",
    "\n",
    "for t in range(mod.nobs):\n",
    "    pred1[:, t] = res.predict(start=t, end=t + n_forecasts - 1, dynamic=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Multiple Kalman filters, filter from the beginning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/fulton/projects/statsmodels/statsmodels/tsa/statespace/mlemodel.py:938: RuntimeWarning: invalid value encountered in true_divide\n",
      "  (self.nobs - self.ssm.loglikelihood_burn)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " "
     ]
    }
   ],
   "source": [
    "n_forecasts = 10\n",
    "pred2 = np.zeros((n_forecasts, mod.nobs))\n",
    "\n",
    "for t in range(mod.nobs):\n",
    "    endog_fcast = [np.nan] if t == 0 else endog[:t]\n",
    "    mod_fcast = sm.tsa.UnobservedComponents(endog_fcast, 'llevel', seasonal=4)\n",
    "    res_fcast = mod_fcast.filter(res.params)\n",
    "    pred2[:, t] = res_fcast.forecast(n_forecasts)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Multiple Kalman filters, only one step filtering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " "
     ]
    }
   ],
   "source": [
    "n_forecasts = 10\n",
    "pred3 = np.zeros((n_forecasts, mod.nobs))\n",
    "\n",
    "for t in range(mod.nobs):\n",
    "    mod_fcast = sm.tsa.UnobservedComponents([], 'llevel', seasonal=4)\n",
    "    mod_fcast.ssm.initialize_known(res.predicted_state[..., t], res.predicted_state_cov[..., t])\n",
    "    res_fcast = mod_fcast.filter(res.params)\n",
    "    pred3[:, t] = res_fcast.forecast(n_forecasts)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test for equality"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "print(np.all(pred1 == pred2))\n",
    "print(np.all(pred2 == pred3))"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}