Changes in version 0.3.0 (2026-07-19)                  

  - Breaking: the quantile-threshold constructor q() is renamed to
    as_quantile() to avoid masking base::q() and for consistency with
    as_traces(). Replace q(0.8) with as_quantile(0.8).
  - Added the maintainer's ORCID.

                        Changes in version 0.2.0                        

Data

  - illustrative_log now ships the authoritative CSV from Delias et al.
    (2023) rather than a reconstruction of Table 1. Breaking: case ids
    are now "1"--"25" (previously G1/B1...), the lower tier is "NORMAL"
    (previously "Blue"), and satisfaction is "High"/"Low" (previously
    upper case). The credibility matrix S is unchanged in value.

Vignettes

  - The reproduction vignettes use the real illustrative log. The
    large-log article now reproduces the BPIC'11 case study of Delias et
    al. (2023), including Figures 2 and 3 from bundled aggregate tables.

Internal

  - Added gated (skip_on_cran) regression tests on the BPIC'11
    similarity matrix. That matrix and the preprocessed BPIC'11 log are
    kept local-only under the 4TU General Terms of Use and are not
    redistributed with the package.

                        Changes in version 0.1.0                        

First public release. simOutrank implements outranking-based trace
clustering for process mining.

Pipeline

  - as_traces() turns a data.frame event log (or, with \pkg{bupaR}, an
    eventlog/activitylog) into a traces object.
  - Criterion helpers build the pairwise measures: process-aware
    (crit_activity_profile(), crit_transitions(), crit_edit_distance(),
    crit_trace_length(), crit_distinct_activities(), crit_duration()),
    attribute templates (crit_nominal(), crit_ordinal(), crit_numeric())
    and the crit_custom() escape hatch, all built on criterion() with
    fixed-number or quantile (as_quantile()) thresholds.
  - outrank_similarity() aggregates the criteria into the ELECTRE-III
    credibility matrix S.
  - cluster_traces() partitions S by normalized spectral clustering
    (Ng–Jordan–Weiss) or hierarchical clustering; eigengap() helps
    choose k.

Robustness and diagnostics

  - trim_outliers() (greedy and exact integer-program methods).
  - must_link() / cannot_link() pairwise constraints.
  - validate_clusters() (connectivity and Dunn index via \pkg{clValid}),
    augment_log() and plot() methods.

Data and documentation

  - Bundled illustrative_log dataset (Delias et al., 2023, Table 1).
  - Vignettes: getting started, reproducing the illustrative example,
    robustness, and parameter tuning; plus website articles reproducing
    the emergency-room study and clustering a large log.