Classification research / edition 2026.10
Compare classifiers
Inspect thresholds, clustering, Markov models and the information set behind a classification.
Threshold Classification
Build an auditable classification specification by declaring features, comparisons, equality behavior, uncertainty handling, and version boundaries.
Markov Switching and State Probabilities
Understand latent-state models, probability information sets, transition assumptions, and why a fitted state path is not decision-time validation.
Clustering and Label Stability
Distinguish a change in cluster identifiers from a change in membership, and assess geometry, time stability, and interpretation separately.
Rolling-Window Standardisation
Interpret standardized features relative to an explicit information-eligible window, with worked calculations showing why endpoint choices matter.
Comparing Regime Classifications
Compare methods on the same eligible information, explain disagreements and avoid confusing agreement with accuracy. A useful comparison reveals rule sensitivity without inventing a true regime.