Add train and test example frequency [Resolves #148]

dssg · May 2, 2017 · b47bb4c · b47bb4c
1 parent 84ac133
commit b47bb4c
Show file tree

Hide file tree

Showing 3 changed files with 13 additions and 5 deletions.
diff --git a/example_experiment_config.yaml b/example_experiment_config.yaml
@@ -11,12 +11,14 @@ temporal_config:
     beginning_of_time: '1995-01-01' # earliest date included in features
     modeling_start_time: '2012-01-01' # earliest date in any model
     modeling_end_time: '2015-01-01' # all dates in any model are < this date
-    update_window: '6m' # how frequently to retrain models
+    update_window: '6month' # how frequently to retrain models
     look_back_durations: # length of time included in a model
-        - '6m'
-        - '3m'
-    test_durations: ['1m', '2m']
-    prediction_frequency: '1d' # per test set how often to generate predictions
+        - '6month'
+        - '3month'
+    train_example_frequency: '1day' # time between rows for same entity in train matrix
+    test_example_frequency: '3month' # time between rows for same entity in test matrix
+    test_durations: ['1month', '2month']
+    prediction_frequency: '1day' # per test set how often to generate predictions
 
 # LABEL GENERATION
 # Information needed to generate labels

diff --git a/tests/test_pipeline.py b/tests/test_pipeline.py
@@ -113,6 +113,8 @@ def simple_pipeline_test(pipeline_class):
             'update_window': '1y',
             'prediction_window': '6months',
             'look_back_durations': ['6months'],
+            'train_example_frequency': '1day',
+            'test_example_frequency': '3months',
             'test_durations': ['1months'],
             'prediction_frequency': '1d'
         }
@@ -241,6 +243,8 @@ def reuse_pipeline_test(pipeline_class):
             'update_window': '1y',
             'prediction_window': '6months',
             'look_back_durations': ['6months'],
+            'train_example_frequency': '1day',
+            'test_example_frequency': '3months',
             'test_durations': ['1months'],
             'prediction_frequency': '1d'
         }

diff --git a/triage/pipelines/base.py b/triage/pipelines/base.py
@@ -69,6 +69,8 @@ def initialize_factories(self):
             modeling_end_time=dt_from_str(split_config['modeling_end_time']),
             update_window=split_config['update_window'],
             look_back_durations=split_config['look_back_durations'],
+            train_example_frequency=split_config['train_example_frequency'],
+            test_example_frequency=split_config['test_example_frequency'],
             test_durations=split_config['test_durations'],
         )