aggregation_functions.md

Window Aggregations

FeatHub can use OverWindowTransform and SlidingWindowTransform to describe how to derive a feature value by applying FeatHub expression and aggregation function on multiple rows. In the following, we provide example usages of these window transforms and the supported built-in aggregation functions.

OverWindowTransform

OverWindowTransform derives a feature value by applying FeatHub expression and aggregation function on multiple rows of a table at a time. It can be used in DerivedFeatureView.

Below is an example usage of OverWindowTransform.

f_total_cost = Feature(
    name="total_cost",
    transform=OverWindowTransform(
        expr="cost",
        agg_func="SUM",
        group_by_keys=["name"],
        window_size=timedelta(days=2),
    ),
)

features = DerivedFeatureView(
    name="feature_view",
    source=source,
    features=[
        f_total_cost,
    ],
    keep_source_fields=False,
)

SlidingWindowTransform

SlidingWindowTransform derives a feature value by applying FeatHub expression and aggregation function on multiple rows in a sliding window. It can be used in SlidingFeatureView.

Below is an example usage of SlidingWindowTransform.

f_total_cost = Feature(
    name="total_cost",
    transform=SlidingWindowTransform(
        expr="cost",
        agg_func="SUM",
        window_size=timedelta(days=3),
        group_by_keys=["name"],
        limit=2,
        step_size=timedelta(days=1),
    ),
)

features = SlidingFeatureView(
  name="features",
  source=source,
  features=[f_total_cost],
)

Built-in Aggregation Functions

The following built-in aggregation functions are supported by the window transforms described above.

Function	Description
AVG	Returns the average (arithmetic mean) of input values.
SUM	Returns the sum of input values.
MAX	Returns the maximum value of input values.
MIN	Returns the minimum value of input values.
FIRST_VALUE	Returns the first value in the ordered list of input values.
LAST_VALUE	Returns the last value in the ordered list of input values.
ROW_NUMBER	Assigns a unique, sequential number to each value in the ordered list of input values, starting with one.
COUNT	Returns the number of input values.
VALUE_COUNTS	Returns a map that maps each value to the number of occurrences of this value in the input values.
COLLECT_LIST	returns a list that contains the ordered list of input values.