Feature release with fixes and improvements across Stochastic Play-out Log Generation, new metric implementations, metric capture in experiments, and analysis scripts.
New Metrics
- Alpha Precision implementations. See Depaire et al (2022) - Alpha Precision: Estimating the Significant System Behavior in a Model
- Existential Precision, a measure inspired by Alpha Precision, but with tolerance of differences across domains / logs. Alpha precision depends on a domain-specific alpha constant.
- Entropic Relevance implementations. See Alkhammash et al (2022) - Entropic relevance: A mechanism for measuring stochastic process models discovered from event data. All three variants are implemented. The SDFA constraint on input models is relaxed by using stochastic playout logs
- Fix rare bug in Generalization by Trace Difference which allows negative measure when TraceFreq is scaled down
-
- This can happen when the model playout size greater than log size
Stochastic Play-out Log Generation
- During stochastic playout log generation surplus allocation, sort silent transitions before other transitions, using the
invisible()method on the net transition
Metric Collection
- Collect various new metrics on all experiment runs, including model entity counts, entropic relevance, and existential precision
- Allow genetic miner runs to have observer metrics collected that are not part of the survival function
Experiment Runner
- Allow calculation of some metrics post-hoc, using the run file and model as input
- Change metrics collected during experiments
-
- Removed collection of Trace Probability Mass Overlap (TMO)
-
- Removed collection of Generalization by Trace Floor 1 and 10
Analysis
- New paradigm cases to further illustrate dimensional extremes
- 3D plots and more selective biplots of PCA space in
dimchoice.R
Environment and Lifecycle
- Now requires Java 11 or later
- Remove the deprecated PPT implementation in package
qut.pm.spm.ppt.legacy - Upgrade Earth Movers to latest API