feature-transform · v1.0.0

Exponential Moving Average

Compute an exponentially weighted moving average of a numeric column, optionally partitioned by a group key and ordered by a sortable key. Pure, leakage-safe; the EMA at row t uses only rows ≤ t within its group.

Contract

Input:: Frame
Output:: Frame
Determinism:: pure
Side effects:: none
Leakage window:: 0s

Params (live form)

valueColumn Name of the numeric column to smooth. outputColumn Name of the new column to write the EMA into. Existing values in this column are overwritten. groupColumn Optional column used to partition rows into independent EMA streams (e.g. 'team_id'). Empty string means no grouping (single stream). orderColumn Column whose ascending value defines the order in which rows are smoothed within each group. Sort is stable; existing row order breaks ties. alpha Smoothing factor in (0, 1]. Higher alpha = more weight on the most recent observation. EMA_t = alpha * x_t + (1 - alpha) * EMA_{t-1}. seed How to initialise the EMA on the first observation of a group. 'first': seed with the first value (default). 'zero': seed with 0.

JSON

{
  "valueColumn": "points_scored",
  "outputColumn": "points_ema",
  "groupColumn": "team_id",
  "orderColumn": "game_date",
  "alpha": 0.3,
  "seed": "first"
}

Examples

Smooth a team-level points-per-game series into a slow-moving form indicator.
Smooth a single ordered sequence with stronger reactivity (alpha=0.6).

Try in builder