Merge pull request #632 from Aarhus-Psychiatry-Research/fix_uc_v1

fix: add unique_count aggregation function to v1

src/timeseriesflattener/v1/aggregation_fns.py

@@ -1,7 +1,6 @@
 """Functions for resolving multiple values in a time-series into a single
 value."""
 
-
 from typing import Callable
 
 import catalogue
@@ -59,6 +58,10 @@ def count(grouped_df: DataFrameGroupBy) -> DataFrame:
     return grouped_df.count()
 
 
+def unique_count(grouped_df: DataFrameGroupBy) -> DataFrame:
+    return grouped_df.nunique()
+
+
 def variance(grouped_df: DataFrameGroupBy) -> DataFrame:
     return grouped_df.var()
 

src/timeseriesflattener/v1/tests/test_timeseriesflattener/test_aggregation_fns.py

@@ -1,12 +1,12 @@
 """Tests of aggregation strategies."""
 
-
 import numpy as np
 
 from timeseriesflattener.v1.aggregation_fns import (
     boolean,
     change_per_day,
     count,
+    unique_count,
     earliest,
     latest,
     maximum,
@@ -237,6 +237,29 @@ def test_aggregation_count():
     )
 
 
+def test_aggregation_unique_count():
+    prediction_times_str = """entity_id,timestamp,
+                            1,2021-12-31 00:00:00
+                            """
+    predictor_df_str = """entity_id,timestamp,value,
+                        1,2021-12-30 00:00:01, 1
+                        1,2021-12-30 00:00:02, 2,
+                        1,2021-12-30 00:00:03, 2,
+                        """
+
+    assert_flattened_data_as_expected(
+        prediction_times_df=prediction_times_str,
+        output_spec=PredictorSpec(
+            feature_base_name="value",
+            timeseries_df=str_to_df(predictor_df_str),
+            aggregation_fn=unique_count,
+            lookbehind_days=2,
+            fallback=0,
+        ),
+        expected_values=[2],
+    )
+
+
 def test_aggregation_bool():
     prediction_times_str = """entity_id,timestamp,
                             1,2021-12-31 00:00:00