Package index • ssel

Application entry points

Start here for complete modelling runs in consumer projects.

modelPipeline(): Orchestrate the supervised ssel workflow
chainPipeline(): Fit an iterative multi-response companion chain
semiSupervisedPipeline(): Fit a range-ratio pseudo-label promotion loop
postProcess(): Assemble user-facing prediction deliverables.

Dataset assembly

buildDataset(): Build per-response training and prediction files

Supervised workflow components

Components orchestrated by modelPipeline for phase-level use.

trainModel(): Fit every discovered response–dataset training cell
predictModel(): Evaluate fitted cells and select response-level deliveries
aggregateResponses(): Select and assemble final response deliveries

Cell-level engine

Advanced fitting and evaluation engine called by trainModel, predictModel, and semiSupervisedPipeline.

trainRegressionModel(): Fit or evaluate a requested response–dataset grid

OOF reconstruction and diagnostics

oofEnsemble(): Reconstruct a projected OOF ensemble
auditQuantiles(): Estimate signed OOF-residual offsets
auditOverfit(): Audit package-defined in-sample optimism
detectOutliers(): Flag regression training outliers

Chain feature gates

extractChainImportance(): Extract normalized importance for response companions
activeByImportance(): Select chain responses by mean normalized importance
activeByShadow(): Select chain responses by a shadow comparison
computeActiveByImportance(): Extract, threshold, and intersect fixed-gate chain responses

Utilities

toNumeric(): Convert supported numeric strings to numeric values
which.nonnum(): Locate values outside base R's numeric grammar
removeOutliersIQR(): Remove rows outside feature IQR envelopes