nirs4all.data.binning module

Binning utilities for regression target values.

This module provides utilities to bin continuous regression targets into discrete classes for balanced augmentation.

class nirs4all.data.binning.BinningCalculator

Bases: object

Calculate bins for continuous regression targets.

static bin_continuous_targets(y: ndarray, bins: int = 10, strategy: str = 'equal_width') → Tuple[ndarray, ndarray]

Bin continuous target values into discrete classes.

Parameters:

• y – Continuous target values (1D array)

• bins – Number of bins (1-1000). Default: 10

• strategy – “quantile” (equal probability) or “equal_width” (uniform spacing, default)

Returns:

Tuple of (bin_indices, bin_edges) - bin_indices: 0-based bin index for each sample - bin_edges: Edge values defining bin boundaries

Raises:

ValueError – If invalid parameters or y contains NaN