feat: Forward method TBATS/AutoTBATS

README.md

@@ -115,6 +115,7 @@ Automatic forecasting tools search for the best parameters and select the best p
 |[AutoETS](https://nixtla.github.io/statsforecast/src/core/models.html#autoets)|✅|✅|✅|✅||
 |[AutoCES](https://nixtla.github.io/statsforecast/src/core/models.html#autoces)|✅|✅|✅|✅||
 |[AutoTheta](https://nixtla.github.io/statsforecast/src/core/models.html#autotheta)|✅|✅|✅|✅||
+|[AutoTBATS](https://nixtla.github.io/statsforecast/src/core/models.html#autotbats)|✅|✅|✅|✅||
 
 ### ARIMA Family
 These models exploit the existing autocorrelations in the time series.
@@ -140,6 +141,7 @@ Suited for signals with more than one clear seasonality. Useful for low-frequenc
 |Model | Point Forecast | Probabilistic Forecast | Insample fitted values | Probabilistic fitted values |Exogenous features|
 |:------|:-------------:|:----------------------:|:---------------------:|:----------------------------:|:----------------:|
 |[MSTL](https://nixtla.github.io/statsforecast/src/core/models.html#mstl)|✅|✅|✅|✅|If trend forecaster supports|
+|[TBATS](https://nixtla.github.io/statsforecast/src/core/models.html#tbats)|✅|✅|✅|✅||
 
 ### GARCH and ARCH Models 
 Suited for modeling time series that exhibit non-constant volatility over time. The ARCH model is a particular case of GARCH. 

nbs/src/core/models.ipynb

@@ -99,7 +99,12 @@
     "from statsforecast.garch import (\n",
     "    garch_model, garch_forecast\n",
     ")\n",
-    "from statsforecast.tbats import tbats_selection, tbats_forecast, _compute_sigmah\n",
+    "from statsforecast.tbats import (\n",
+    "    tbats_selection, \n",
+    "    tbats_forecast, \n",
+    "    _compute_sigmah, \n",
+    "    forward_tbats\n",
+    ")\n",
     "from statsforecast.utils import (\n",
     "    _calculate_sigma,\n",
     "    _calculate_intervals,\n",
@@ -9953,7 +9958,60 @@
     "            res_trans = {k : inv_boxcox(v, mod['BoxCox_lambda']) for k, v in res.items()}\n",
     "        else:\n",
     "            res_trans = res\n",
-    "        return res_trans "
+    "        return res_trans \n",
+    "    \n",
+    "    def forward(\n",
+    "            self,\n",
+    "            y: np.ndarray,\n",
+    "            h: int,\n",
+    "            X: Optional[np.ndarray] = None,\n",
+    "            X_future: Optional[np.ndarray] = None,\n",
+    "            level: Optional[List[int]] = None,\n",
+    "            fitted: bool = False,\n",
+    "        ):\n",
+    "        \"\"\"Apply fitted TBATS model to a new time series.\n",
+    "\n",
+    "        Parameters\n",
+    "        ----------\n",
+    "        y : numpy.array \n",
+    "            Clean time series of shape (n, ). \n",
+    "        h : int \n",
+    "            Forecast horizon.\n",
+    "        X : array-like \n",
+    "            Optional insample exogenpus of shape (t, n_x). \n",
+    "        X_future : array-like \n",
+    "            Optional exogenous of shape (h, n_x). \n",
+    "        level : List[float] \n",
+    "            Confidence levels for prediction intervals.\n",
+    "        fitted : bool \n",
+    "            Whether or not to return insample predictions.\n",
+    "\n",
+    "        Returns\n",
+    "        -------\n",
+    "        forecasts : dict \n",
+    "            Dictionary with entries `mean` for point predictions and `level_*` for probabilistic predictions.\n",
+    "        \"\"\"\n",
+    "        if not hasattr(self, 'model_'):\n",
+    "            raise Exception('You have to use the `fit` method first')\n",
+    "        mod = forward_tbats(self.model_, y)\n",
+    "        fcst = tbats_forecast(mod, h)\n",
+    "        res = {'mean': fcst['mean']}\n",
+    "        if fitted:\n",
+    "            res['fitted'] = mod['fitted'].ravel()\n",
+    "        if level is not None:\n",
+    "            level = sorted(level)\n",
+    "            sigmah = _compute_sigmah(mod, h)\n",
+    "            pred_int = _calculate_intervals(res, level, h, sigmah)\n",
+    "            res = {**res, **pred_int}\n",
+    "            if fitted:\n",
+    "                se = _calculate_sigma(mod['errors'], mod['errors'].shape[1])\n",
+    "                fitted_pred_int = _add_fitted_pi(res, se, level)\n",
+    "                res = {**res, **fitted_pred_int}\n",
+    "        if mod['BoxCox_lambda'] is not None:\n",
+    "            res_trans = {k : inv_boxcox(v, mod['BoxCox_lambda']) for k, v in res.items()}\n",
+    "        else:\n",
+    "            res_trans = res\n",
+    "        return res_trans"
    ]
   },
   {
@@ -10012,6 +10070,15 @@
     "show_doc(TBATS.forecast, title_level=3)"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "show_doc(TBATS.forward, title_level=3)"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -10115,7 +10182,16 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "show_doc(AutoTBATS.fit, title_level=3)"
+    "show_doc(AutoTBATS.fit, name='AutoTBATS.fit', title_level=3)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "show_doc(AutoTBATS.predict, name='AutoTBATS.predict', title_level=3)"
    ]
   },
   {
@@ -10124,7 +10200,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "show_doc(AutoTBATS.predict, title_level=3)"
+    "show_doc(AutoTBATS.predict_in_sample, name='AutoTBATS.predict_in_sample', title_level=3)"
    ]
   },
   {
@@ -10133,7 +10209,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "show_doc(AutoTBATS.predict_in_sample, title_level=3)"
+    "show_doc(AutoTBATS.forecast, name='AutoTBATS.forecast', title_level=3)"
    ]
   },
   {
@@ -10142,7 +10218,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "show_doc(AutoTBATS.forecast, title_level=3)"
+    "show_doc(AutoTBATS.forward, name='AutoTBATS.forward', title_level=3)"
    ]
   },
   {