Skip to contents

Memory-based learning prediction against the OSSL library

Source: R/spectra-predict.R
predict_ossl_mbl.Rd

Predicts a set of soil properties from pre-processed Vis-NIR or MIR spectra using memory-based learning (MBL) – the recommended OSSL workflow for heterogeneous libraries. Defaults follow the literature (Ramirez-Lopez et al., 2013): k = 100 neighbours, PLS-score dissimilarity, local PLS regression with 5 components, internal leave-one-out validation.

Usage

predict_ossl_mbl(
  X,
  properties,
  region = "global",
  k = 100L,
  ossl_library = NULL,
  ...
)

Arguments

X

A pre-processed numeric matrix (rows = horizons, columns = wavelengths).

properties

Character vector of OSSL-supported property names.

region

One of "global", "south_america", "north_america", "europe", "africa".

k

Integer number of neighbours.

ossl_library

Optional list with the OSSL training spectra (Xr) and reference values (Yr, a data.frame keyed by properties). When NULL, the synthetic path is used.

...

Additional arguments forwarded to resemble::mbl.

Value

A data.table with columns horizon_idx, property, value, pi95_low, pi95_high, n_neighbors. The "backend" attribute records which path was taken ("resemble" or "synthetic").

Details

If resemble::mbl is installed and an ossl_library artefact is supplied (a list with elements Xr, Yr) the function delegates to resemble::mbl(); otherwise it returns a deterministic synthetic prediction conditioned on the input spectra so that downstream code, tests and vignettes run without external dependencies. The fallback is annotated via the notes attribute on the returned data.table.

References

Ramirez-Lopez, L., Behrens, T., Schmidt, K., Stevens, A., Demattê, J. A. M., & Scholten, T. (2013). The spectrum-based learner: A new local approach for modeling soil Vis-NIR spectra of complex datasets. Geoderma, 195–196, 268–279.