Use fit kwargs in cross validation (#1040)

R/R/diagnostics.R

@@ -117,7 +117,8 @@ cross_validation <- function(
     if (nrow(history.c) < 2) {
       stop('Less than two datapoints before cutoff. Increase initial window.')
     }
-    m <- fit.prophet(m, history.c)
+    fit.args <- c(list(m=m, df=history.c), model$fit.kwargs)
+    m <- do.call(fit.prophet, fit.args)
     # Calculate yhat
     df.predict <- dplyr::filter(df, ds > cutoff, ds <= cutoff + horizon.dt)
     # Get the columns for the future dataframe

R/R/prophet.R

@@ -134,7 +134,8 @@ prophet <- function(df = NULL,
     history.dates = NULL,
     train.holiday.names = NULL,
     train.component.cols = NULL,
-    component.modes = NULL
+    component.modes = NULL,
+    fit.kwargs = list()
   )
   m <- validate_inputs(m)
   class(m) <- append("prophet", class(m))
@@ -1208,6 +1209,7 @@ fit.prophet <- function(m, df, ...) {
   component.cols <- out2$component.cols
   m$train.component.cols <- component.cols
   m$component.modes <- out2$modes
+  m$fit.kwargs <- list(...)
 
   m <- set_changepoints(m)
 

R/tests/testthat/test_prophet.R

@@ -114,6 +114,7 @@ test_that("logistic_floor", {
   expect_true(m$logistic.floor)
   expect_true('floor' %in% colnames(m$history))
   expect_equal(m$history$y_scaled[1], 1., tolerance = 1e-6)
+  expect_equal(m$fit.kwargs, list(algorithm = 'Newton'))
   fcst1 <- predict(m, future1)
 
   m2 <- prophet(growth = 'logistic')

python/fbprophet/diagnostics.py

@@ -117,7 +117,7 @@ def cross_validation(model, horizon, period=None, initial=None):
                 'Less than two datapoints before cutoff. '
                 'Increase initial window.'
             )
-        m.fit(history_c)
+        m.fit(history_c, **model.fit_kwargs)
         # Calculate yhat
         index_predicted = (df['ds'] > cutoff) & (df['ds'] <= cutoff + horizon)
         # Get the columns for the future dataframe

python/fbprophet/forecaster.py

@@ -8,6 +8,7 @@ from __future__ import absolute_import, division, print_function
 
 import logging
 from collections import OrderedDict, defaultdict
+from copy import deepcopy
 from datetime import timedelta
 
 import numpy as np
@@ -137,6 +138,7 @@ class Prophet(object):
         self.train_component_cols = None
         self.component_modes = None
         self.train_holiday_names = None
+        self.fit_kwargs = {}
         self.validate_inputs()
 
     def validate_inputs(self):
@@ -1080,6 +1082,7 @@ class Prophet(object):
             self.make_all_seasonality_features(history))
         self.train_component_cols = component_cols
         self.component_modes = modes
+        self.fit_kwargs = deepcopy(kwargs)
 
         self.set_changepoints()
 

python/fbprophet/tests/test_diagnostics.py

@@ -109,7 +109,7 @@ class TestDiagnostics(TestCase):
         df = self.__df.copy()
         for uncertainty in [0, False]:
             m = Prophet(uncertainty_samples=uncertainty)
-            m.fit(df)
+            m.fit(df, algorithm='Newton')
             df_cv = diagnostics.cross_validation(
                 m, horizon='4 days', period='4 days', initial='115 days')
             expected_cols = ['ds', 'yhat', 'y', 'cutoff']

python/fbprophet/tests/test_prophet.py

@@ -142,6 +142,7 @@ class TestProphet(TestCase):
         self.assertTrue(m.logistic_floor)
         self.assertTrue('floor' in m.history)
         self.assertAlmostEqual(m.history['y_scaled'][0], 1.)
+        self.assertEqual(m.fit_kwargs, {'algorithm': 'Newton'})
         fcst1 = m.predict(future)
 
         m2 = Prophet(growth='logistic')