diff --git a/tsforecast.ipynb b/tsforecast.ipynb index f214687..319eb48 100644 --- a/tsforecast.ipynb +++ b/tsforecast.ipynb @@ -1425,6 +1425,9 @@ " method=\"nm\",\n", " maxiter=100\n", ")\n", + "#TODO : Try out different values for m (seasonality) -- can influence comp. cost\n", + "#TODO : Perform feature engineering, e.g. rolling means\n", + "#TODO : Try out different history sizes\n", "\n", "# Forecast\n", "y_pred = autoarima_model.predict(n_periods=len(y_test))\n", @@ -1470,7 +1473,7 @@ }, { "cell_type": "code", - "execution_count": 38, + "execution_count": null, "id": "9796045c", "metadata": {}, "outputs": [ @@ -1503,6 +1506,9 @@ " hist_size=hist_size\n", ")\n", "\n", + "#TODO : Perform feature engineering, e.g. rolling means\n", + "#TODO : Perform feature selection and combine with e.g. PCA\n", + "#TODO : Try out different history sizes\n", "\n", "X_train_xgb = reshaping(X_train)\n", "X_test_xgb = reshaping(X_test)\n", @@ -38817,7 +38823,7 @@ }, { "cell_type": "code", - "execution_count": 41, + "execution_count": null, "id": "abeddbe8", "metadata": {}, "outputs": [ @@ -39034,7 +39040,8 @@ "model.add(Dense(1))\n", "model.compile(optimizer=\"adam\", loss=\"mse\")\n", "model.summary()\n", - "\n", + "#TODO : Try out different architectures, like Conv1D, Bidirectional LSTM, etc.\n", + "#TODO : Try out different history sizes\n", "\n", "# fit the model\n", "model.fit(X_train, y_train, epochs=50)\n", @@ -40432,6 +40439,8 @@ "\n", "# Initialize and fit the Prophet model\n", "prophet_model = prophet.Prophet()\n", + "#TODO : Add any additional Prophet parameters if needed\n", + "#TODO : Add extra regressors if needed\n", "prophet_model.fit(df_train)\n", "\n", "# Create future dataframe for prediction\n", @@ -40481,8 +40490,7 @@ "axes[1].legend()\n", "\n", "plt.tight_layout()\n", - "plt.show()\n", - "\n" + "plt.show()\n" ] } ],