Merge pull request #241 from winedarksea/dev

0.6.12

TODO.md

@@ -13,12 +13,14 @@
 * Forecasts are desired for the future immediately following the most recent data.
 * trimmed_mean to AverageValueNaive
 
-# 0.6.11 🇺🇦 🇺🇦 🇺🇦
+# 0.6.12 🇺🇦 🇺🇦 🇺🇦
 * bug fixes
-* continually trying to keep up with the Pandas maintainers who are breaking stuff for no good reasonable
-* updated RollingMeanTransformer and RegressionFilter, RegressionFilter should now be less memory intensive
-* EIA data call to load_live_daily
-* horizontal_ensemble_validation arg for more complete validation on these ensembles
+* added DMD model
+* modified the `constraints` options so it now accepts of list of dictionaries of constraints with new last_window and slope options
+* 'dampening' as a constraint method to dampen all forecasts, fixed Cassandra trend_phi dampening
+* new med_diff anomaly method and 'laplace' added as distribution option
+* modified fourier_df to now work with sub daily data
+* some madness with wavelets attempting to use them like fourier series for seasonality
 
 ### Unstable Upstream Pacakges (those that are frequently broken by maintainers)
 * Pytorch-Forecasting

autots/__init__.py

@@ -27,7 +27,7 @@
 from autots.models.cassandra import Cassandra
 
 
-__version__ = '0.6.11'
+__version__ = '0.6.12'
 
 TransformTS = GeneralTransformer
 

autots/evaluator/auto_model.py

@@ -61,7 +61,7 @@
     DynamicFactorMQ,
 )
 from autots.models.arch import ARCH
-from autots.models.matrix_var import RRVAR, MAR, TMF, LATC
+from autots.models.matrix_var import RRVAR, MAR, TMF, LATC, DMD
 
 
 def create_model_id(
@@ -698,6 +698,17 @@ def ModelMonster(
             n_jobs=n_jobs,
             **parameters,
         )
+    elif model == 'DMD':
+        return DMD(
+            frequency=frequency,
+            prediction_interval=prediction_interval,
+            holiday_country=holiday_country,
+            random_seed=random_seed,
+            verbose=verbose,
+            forecast_length=forecast_length,
+            n_jobs=n_jobs,
+            **parameters,
+        )
     elif model == "":
         raise AttributeError(
             ("Model name is empty. Likely this means AutoTS has not been fit.")
@@ -864,7 +875,7 @@ def predict(self, forecast_length=None, future_regressor=None):
         if not self._fit_complete:
             raise ValueError("Model not yet fit.")
         df_forecast = self.model.predict(
-            forecast_length=self.forecast_length, future_regressor=future_regressor
+            forecast_length=forecast_length, future_regressor=future_regressor
         )
 
         # THIS CHECKS POINT FORECAST FOR NULLS BUT NOT UPPER/LOWER FORECASTS
@@ -896,11 +907,13 @@ def predict(self, forecast_length=None, future_regressor=None):
         # CHECK Forecasts are proper length!
         if df_forecast.forecast.shape[0] != self.forecast_length:
             raise ValueError(
-                f"Model {self.model_str} returned improper forecast_length"
+                f"Model {self.model_str} returned improper forecast_length. Returned: {df_forecast.forecast.shape[0]} and requested: {self.forecast_length}"
             )
 
         if df_forecast.forecast.shape[1] != self.df.shape[1]:
-            raise ValueError("Model failed to return correct number of series.")
+            raise ValueError(
+                f"Model failed to return correct number of series. Returned {df_forecast.forecast.shape[1]} and requested: {self.df.shape[1]}"
+            )
 
         df_forecast.transformation_parameters = self.transformation_dict
         # Remove negatives if desired
@@ -911,33 +924,53 @@ def predict(self, forecast_length=None, future_regressor=None):
             df_forecast.upper_forecast = df_forecast.upper_forecast.clip(lower=0)
 
         if self.constraint is not None:
-            if isinstance(self.constraint, dict):
-                constraint_method = self.constraint.get("constraint_method", "quantile")
-                constraint_regularization = self.constraint.get(
-                    "constraint_regularization", 1
+            if isinstance(self.constraint, list):
+                constraints = self.constraint
+                df_forecast = df_forecast.apply_constraints(
+                    constraints=constraints,
+                    df_train=self.df,
                 )
-                lower_constraint = self.constraint.get("lower_constraint", 0)
-                upper_constraint = self.constraint.get("upper_constraint", 1)
-                bounds = self.constraint.get("bounds", False)
             else:
-                constraint_method = "stdev_min"
-                lower_constraint = float(self.constraint)
-                upper_constraint = float(self.constraint)
-                constraint_regularization = 1
-                bounds = False
-            if self.verbose > 3:
-                print(
-                    f"Using constraint with method: {constraint_method}, {constraint_regularization}, {lower_constraint}, {upper_constraint}, {bounds}"
-                )
+                constraints = None
+                if isinstance(self.constraint, dict):
+                    if "constraints" in self.constraint.keys():
+                        constraints = self.constraint.get("constraints")
+                        constraint_method = None
+                        constraint_regularization = None
+                        lower_constraint = None
+                        upper_constraint = None
+                        bounds = True
+                    else:
+                        constraint_method = self.constraint.get(
+                            "constraint_method", "quantile"
+                        )
+                        constraint_regularization = self.constraint.get(
+                            "constraint_regularization", 1
+                        )
+                        lower_constraint = self.constraint.get("lower_constraint", 0)
+                        upper_constraint = self.constraint.get("upper_constraint", 1)
+                        bounds = self.constraint.get("bounds", False)
+                else:
+                    constraint_method = "stdev_min"
+                    lower_constraint = float(self.constraint)
+                    upper_constraint = float(self.constraint)
+                    constraint_regularization = 1
+                    bounds = False
+                if self.verbose > 3:
+                    print(
+                        f"Using constraint with method: {constraint_method}, {constraint_regularization}, {lower_constraint}, {upper_constraint}, {bounds}"
+                    )
 
-            df_forecast = df_forecast.apply_constraints(
-                constraint_method,
-                constraint_regularization,
-                upper_constraint,
-                lower_constraint,
-                bounds,
-                self.df,
-            )
+                print(constraints)
+                df_forecast = df_forecast.apply_constraints(
+                    constraints,
+                    self.df,
+                    constraint_method,
+                    constraint_regularization,
+                    upper_constraint,
+                    lower_constraint,
+                    bounds,
+                )
 
         self.transformation_runtime = self.transformation_runtime + (
             datetime.datetime.now() - transformationStartTime
@@ -966,6 +999,18 @@ def fit_data(self, df, future_regressor=None):
         self.df = df
         self.model.fit_data(df, future_regressor)
 
+    def fit_predict(
+        self,
+        df,
+        forecast_length,
+        future_regressor_train=None,
+        future_regressor_forecast=None,
+    ):
+        self.fit(df, future_regressor=future_regressor_train)
+        return self.predict(
+            forecast_length=forecast_length, future_regressor=future_regressor_forecast
+        )
+
 
 class TemplateEvalObject(object):
     """Object to contain all the failures!.
@@ -2119,7 +2164,9 @@ def NewGeneticTemplate(
 
     # filter existing templates
     sorted_results = model_results[
-        (model_results['Ensemble'] == 0) & (model_results['Exceptions'].isna())
+        (model_results['Ensemble'] == 0)
+        & (model_results['Exceptions'].isna())
+        & (model_results['Model'].isin(model_list))
     ].copy()
     # remove duplicates by exact same performance
     sorted_results = sorted_results.sort_values(

autots/evaluator/auto_ts.py

@@ -1075,6 +1075,7 @@ def fit_data(
             preclean=None,
             verbose=0,
         )
+        return self
 
     def fit(
         self,
@@ -1826,8 +1827,10 @@ def _run_template(
             self.model_count = template_result.model_count
         # capture results from lower-level template run
         if "TotalRuntime" in template_result.model_results.columns:
-            template_result.model_results['TotalRuntime'].fillna(
-                pd.Timedelta(seconds=60), inplace=True
+            template_result.model_results['TotalRuntime'] = (
+                template_result.model_results['TotalRuntime'].fillna(
+                    pd.Timedelta(seconds=60)
+                )
             )
         else:
             # trying to catch a rare and sneaky bug (perhaps some variety of beetle?)
@@ -2161,9 +2164,13 @@ def results(self, result_set: str = 'initial'):
             result_set (str): 'validation' or 'initial'
         """
         if result_set == 'validation':
-            return self.validation_results.model_results
+            return self.validation_results.model_results.sort_values(
+                "Score", ascending=True
+            )
         else:
-            return self.initial_results.model_results
+            return self.initial_results.model_results.sort_values(
+                "Score", ascending=True
+            )
 
     def failure_rate(self, result_set: str = 'initial'):
         """Return fraction of models passing with exceptions.
@@ -2280,6 +2287,22 @@ def export_template(
             export_template = unpack_ensemble_models(
                 export_template, self.template_cols, keep_ensemble=False, recursive=True
             ).drop_duplicates()
+            if include_results:
+                export_template = export_template.drop(columns=['smape']).merge(
+                    self.validation_results.model_results[['ID', 'smape']],
+                    on="ID",
+                    how='left',
+                )
+                # put smape back in the front
+                remaining_columns = [
+                    col
+                    for col in export_template.columns
+                    if col not in self.template_cols_id and col not in ['smape', 'Runs']
+                ]
+                new_order = (
+                    self.template_cols_id + ['Runs', 'smape'] + remaining_columns
+                )
+                export_template = export_template.reindex(columns=new_order)
         return self.save_template(filename, export_template)
 
     def save_template(self, filename, export_template, **kwargs):