Merge pull request #18 from vonum/cross-validation

Cross validation and performance metrics integration
vonum · May 21, 2021 · 7af3dd4 · 7af3dd4
2 parents e8966fb + e3eb0ea
commit 7af3dd4
Show file tree

Hide file tree

Showing 6 changed files with 78 additions and 0 deletions.
diff --git a/multi_prophet/__init__.py b/multi_prophet/__init__.py
@@ -62,6 +62,18 @@ def plot_components(self, forecasts, plotly=False, **kwargs):
             for column, forecast in forecasts.items()
         }
 
+    def cross_validation(self, horizon, **kwargs):
+        return {
+            column: model.cross_validation(horizon=horizon, **kwargs)
+            for column, model in self.model_pool.items()
+        }
+
+    def performance_metrics(self, horizon, **kwargs):
+        return {
+            column: model.performance_metrics(horizon=horizon, **kwargs)
+            for column, model in self.model_pool.items()
+        }
+
     def _init_model_pool(self, columns, **kwargs):
         return {c: Prophet(**kwargs) for c in columns}
 

diff --git a/multi_prophet/factories.py b/multi_prophet/factories.py
@@ -1,6 +1,7 @@
 from .prophet import Prophet
 from .data_builder import DataFrameBuilder
 
+
 def model_pool_factory(columns=None, config=None, regressors={}, **kwargs):
     if config:
         return _different_models_pool_factory(config, regressors)

diff --git a/multi_prophet/plots.py b/multi_prophet/plots.py
@@ -1,5 +1,6 @@
 from fbprophet.plot import plot_plotly, plot_components_plotly
 
+
 def plotly_plot(model, forecast, **kwargs):
     return plot_plotly(model, forecast, **kwargs)
 

diff --git a/multi_prophet/prophet.py b/multi_prophet/prophet.py
@@ -1,4 +1,5 @@
 import fbprophet
+from fbprophet.diagnostics import cross_validation, performance_metrics
 from . import plots
 
 
@@ -36,3 +37,10 @@ def plot_components(self, forecast, plotly=False, **kwargs):
             return plots.plotly_components_plot(self.prophet, forecast, **kwargs)
         else:
             return self.prophet.plot_components(forecast)
+
+    def cross_validation(self, horizon, **kwargs):
+        return cross_validation(self.prophet, horizon=horizon, **kwargs)
+
+    def performance_metrics(self, horizon, **kwargs):
+        cv_df = cross_validation(self.prophet, horizon=horizon, **kwargs)
+        return performance_metrics(cv_df)
diff --git a/tests/test_multi_prophet.py b/tests/test_multi_prophet.py
@@ -159,3 +159,30 @@ def test_components_plot(self):
 
         for plot in plots:
           self.assertIsInstance(plot, matplotlib.figure.Figure)
+
+    def test_cross_validation(self):
+        mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
+        mp.fit(self.df)
+
+        cross_validation_dfs = mp.cross_validation(horizon="365 days")
+        for c, cross_validation_df in cross_validation_dfs.items():
+            self.assertTrue("ds" in cross_validation_df.columns)
+            self.assertTrue("yhat" in cross_validation_df.columns)
+            self.assertTrue("yhat_lower" in cross_validation_df.columns)
+            self.assertTrue("yhat_upper" in cross_validation_df.columns)
+            self.assertTrue("y" in cross_validation_df.columns)
+            self.assertTrue("cutoff" in cross_validation_df.columns)
+
+    def test_performance_metrics(self):
+        mp = multi_prophet.MultiProphet(columns=PREDICTOR_COLUMNS)
+        mp.fit(self.df)
+
+        performance_metrics_dfs = mp.performance_metrics(horizon="365 days")
+        for c, performance_metrics_df in performance_metrics_dfs.items():
+            self.assertTrue("horizon" in performance_metrics_df.columns)
+            self.assertTrue("mse" in performance_metrics_df.columns)
+            self.assertTrue("rmse" in performance_metrics_df.columns)
+            self.assertTrue("mae" in performance_metrics_df.columns)
+            self.assertTrue("mape" in performance_metrics_df.columns)
+            self.assertTrue("mdape" in performance_metrics_df.columns)
+            self.assertTrue("coverage" in performance_metrics_df.columns)
diff --git a/tests/test_prophet.py b/tests/test_prophet.py
@@ -166,3 +166,32 @@ def test_plotly_components_plot(self):
             mp.plot_components(forecast, plotly=True),
             plotly.graph_objs.Figure
         )
+
+    def test_cross_validation(self):
+        mp = multi_prophet.Prophet()
+        mp.fit(self.df)
+
+        cross_validation_df = mp.cross_validation(horizon="365 days")
+        self.assertIsInstance(cross_validation_df, pd.DataFrame)
+
+        self.assertTrue("ds" in cross_validation_df.columns)
+        self.assertTrue("yhat" in cross_validation_df.columns)
+        self.assertTrue("yhat_lower" in cross_validation_df.columns)
+        self.assertTrue("yhat_upper" in cross_validation_df.columns)
+        self.assertTrue("y" in cross_validation_df.columns)
+        self.assertTrue("cutoff" in cross_validation_df.columns)
+
+    def test_performance_metrics(self):
+        mp = multi_prophet.Prophet()
+        mp.fit(self.df)
+
+        performance_metrics_df = mp.performance_metrics(horizon="365 days")
+        self.assertIsInstance(performance_metrics_df, pd.DataFrame)
+
+        self.assertTrue("horizon" in performance_metrics_df.columns)
+        self.assertTrue("mse" in performance_metrics_df.columns)
+        self.assertTrue("rmse" in performance_metrics_df.columns)
+        self.assertTrue("mae" in performance_metrics_df.columns)
+        self.assertTrue("mape" in performance_metrics_df.columns)
+        self.assertTrue("mdape" in performance_metrics_df.columns)
+        self.assertTrue("coverage" in performance_metrics_df.columns)