`knn` – k-nearest-neighbours regression.#

Operational op under axis family, sub-layer L4_A_model_selection, layer l4. Standalone callable: mf.functions.knn_fit.

Function signature#

mf.functions.knn_fit(
    X: np.ndarray | pd.DataFrame,
    y: np.ndarray | pd.Series,
) -> KNNFitResult

name	type	default	constraint	description
`X`	`np.ndarray	pd.DataFrame`	—	—
`y`	`np.ndarray	pd.Series`	—	—

KNNFitResult — frozen dataclass with fit results.

Attribute	Type	Description
`.n_neighbors`	`int`	Neighbour count k requested.
`.n_neighbors_used`	`int`	Actual k used (clipped to training set size).
`.n_features_in_`	`int`	Number of input features.
`.predict(X)`	`np.ndarray`	Predictions for new data X, shape (n_samples,).
`.summary()`	`str`	Table: neighbour counts and feature count.

Memorises training data; predicts via nearest-neighbour averaging. Cheap, non-parametric.

When to use

Non-parametric baselines; sensitivity studies.

Set params.family = "knn" in the relevant layer to activate this op within a recipe:

# Layer L4 recipe fragment
params:
  family: knn

macroforecast design Part 2, L4: ‘forecasting model is the layer where every authoring iteration ends – pick family, tune, repeat.’
Cover & Hart (1967) ‘Nearest neighbor pattern classification’, IEEE Trans. on Information Theory 13(1).