Add remove_alts option for model groups

This controls whether alternatives are removed from the alternatives
pool between doing prediction for different segments.
When doing LCMs (e.g. probability_mode='single_chooser'
and choice_mode='aggregate') this should be set to True.
For doing something like automobile ownership this should be False.
False is the default.

urbansim/models/dcm.py

@@ -776,13 +776,22 @@ class MNLDiscreteChoiceModelGroup(DiscreteChoiceModel):
 
     Parameters
     ----------
-    segmentation_col
+    segmentation_col : str
         Name of a column in the table of choosers. Will be used to perform
         a pandas groupby on the choosers table.
+    remove_alts : bool, optional
+        Specify how to handle alternatives between prediction for different
+        models. If False, the alternatives table is not modified between
+        predictions. If True, alternatives that have been chosen
+        are removed from the alternatives table before doing another
+        round of prediction.
+    name : str, optional
+        A name that may be used in places to identify this group.
 
     """
-    def __init__(self, segmentation_col, name=None):
+    def __init__(self, segmentation_col, remove_alts=False, name=None):
         self.segmentation_col = segmentation_col
+        self.remove_alts = remove_alts
         self.name = name if name is not None else 'MNLDiscreteChoiceModelGroup'
         self.models = {}
 
@@ -1088,6 +1097,9 @@ def predict(self, choosers, alternatives, debug=False):
 
         for name, df in self._iter_groups(choosers):
             choices = self.models[name].predict(df, alternatives, debug=debug)
+            if self.remove_alts:
+                alternatives = alternatives.loc[
+                    ~alternatives.index.isin(choices)]
             results.append(choices)
 
         logger.debug(
@@ -1177,6 +1189,12 @@ class SegmentedMNLDiscreteChoiceModel(DiscreteChoiceModel):
         the alternatives index is used.
     default_model_expr : str, iterable, or dict, optional
         A patsy model expression. Should contain only a right-hand side.
+    remove_alts : bool, optional
+        Specify how to handle alternatives between prediction for different
+        models. If False, the alternatives table is not modified between
+        predictions. If True, alternatives that have been chosen
+        are removed from the alternatives table before doing another
+        round of prediction.
     name : str, optional
         An optional string used to identify the model in places.
 
@@ -1188,7 +1206,8 @@ def __init__(
             alts_fit_filters=None, alts_predict_filters=None,
             interaction_predict_filters=None,
             estimation_sample_size=None,
-            choice_column=None, default_model_expr=None, name=None):
+            choice_column=None, default_model_expr=None, remove_alts=False,
+            name=None):
         self.segmentation_col = segmentation_col
         self.sample_size = sample_size
         self.probability_mode = probability_mode
@@ -1201,7 +1220,9 @@ def __init__(
         self.estimation_sample_size = estimation_sample_size
         self.choice_column = choice_column
         self.default_model_expr = default_model_expr
-        self._group = MNLDiscreteChoiceModelGroup(segmentation_col)
+        self.remove_alts = remove_alts
+        self._group = MNLDiscreteChoiceModelGroup(
+            segmentation_col, remove_alts=remove_alts)
         self.name = (name if name is not None else
                      'SegmentedMNLDiscreteChoiceModel')
 
@@ -1240,6 +1261,7 @@ def from_yaml(cls, yaml_str=None, str_or_buffer=None):
             cfg['estimation_sample_size'],
             cfg['choice_column'],
             default_model_expr,
+            cfg['remove_alts'],
             cfg['name'])
 
         if "models" not in cfg:
@@ -1565,6 +1587,7 @@ def to_dict(self):
             'default_config': {
                 'model_expression': self.default_model_expr,
             },
+            'remove_alts': self.remove_alts,
             'fitted': self.fitted,
             'models': {
                 yamlio.to_scalar_safe(name):

urbansim/models/tests/test_dcm.py

@@ -258,7 +258,7 @@ def test_mnl_dcm_choice_mode_agg(seed, basic_dcm_fit, choosers, alternatives):
         pd.Series(['f', 'a', 'd', 'c'], index=[0, 1, 3, 4]))
 
 
-def test_mnl_dcm_group(grouped_choosers, alternatives):
+def test_mnl_dcm_group(seed, grouped_choosers, alternatives):
     model_exp = 'var2 + var1:var3'
     sample_size = 4
     choosers_predict_filters = ['var1 != 7']
@@ -296,10 +296,22 @@ def test_mnl_dcm_group(grouped_choosers, alternatives):
     sprobs = group.summed_probabilities(grouped_choosers, alternatives)
     assert len(sprobs) == len(filtered_alts)
 
+    choice_state = np.random.get_state()
     choices = group.predict(grouped_choosers, alternatives)
 
-    assert len(choices) == len(filtered_choosers)
-    assert choices.isin(alternatives.index).all()
+    pdt.assert_series_equal(
+        choices,
+        pd.Series(['c', 'a', 'a', 'g'], index=[0, 3, 1, 4]))
+
+    # check that we don't get the same alt twice if they are removed
+    # make sure we're starting from the same random state as the last draw
+    np.random.set_state(choice_state)
+    group.remove_alts = True
+    choices = group.predict(grouped_choosers, alternatives)
+
+    pdt.assert_series_equal(
+        choices,
+        pd.Series(['c', 'a', 'b', 'g'], index=[0, 3, 1, 4]))
 
 
 def test_mnl_dcm_segmented_raises():
@@ -309,7 +321,7 @@ def test_mnl_dcm_segmented_raises():
         group.add_segment('x')
 
 
-def test_mnl_dcm_segmented(grouped_choosers, alternatives):
+def test_mnl_dcm_segmented(seed, grouped_choosers, alternatives):
     model_exp = 'var2 + var1:var3'
     sample_size = 4
 
@@ -338,10 +350,22 @@ def test_mnl_dcm_segmented(grouped_choosers, alternatives):
     sprobs = group.summed_probabilities(grouped_choosers, alternatives)
     assert len(sprobs) == len(alternatives)
 
+    choice_state = np.random.get_state()
     choices = group.predict(grouped_choosers, alternatives)
 
-    assert len(choices) == len(grouped_choosers)
-    assert choices.isin(alternatives.index).all()
+    pdt.assert_series_equal(
+        choices,
+        pd.Series(['c', 'a', 'b', 'a', 'j'], index=[0, 2, 3, 1, 4]))
+
+    # check that we don't get the same alt twice if they are removed
+    # make sure we're starting from the same random state as the last draw
+    np.random.set_state(choice_state)
+    group._group.remove_alts = True
+    choices = group.predict(grouped_choosers, alternatives)
+
+    pdt.assert_series_equal(
+        choices,
+        pd.Series(['c', 'a', 'b', 'd', 'j'], index=[0, 2, 3, 1, 4]))
 
 
 def test_mnl_dcm_segmented_yaml(grouped_choosers, alternatives):
@@ -370,6 +394,7 @@ def test_mnl_dcm_segmented_yaml(grouped_choosers, alternatives):
         'default_config': {
             'model_expression': model_exp,
         },
+        'remove_alts': False,
         'fitted': False,
         'models': {
             'x': {