Discovered Process Models from Noisy Logs

<p dir="ltr">The dataset consists of three folders:</p><ol><li><b>Systems:</b> Three public event logs: <i>Sepsis Cases</i>, <i>RTFMS</i>, and <i>BPIC 2012</i>.</li><li><b>Logs:</b> Clean and noisy logs derived from the base systems.<br>From each base log, we created samples of seven sizes (1000, 2000, 4000, 10000, 20000, 40000, 100000 traces) using sampling with replacement, yielding 21 clean logs.<br>Noise was then added using $\snip$ across seven intensity levels (0.1%, 0.2%, 0.4%, 1.0%, 2.0%, 4.0%, 10.0%) and five noise types (absence, insertion, ordering, substitution, mixed). Percentages refer to the number of trace-level injections.<br>Each configuration was repeated five times, producing 3,675 noisy logs and a total of 3,696 logs.</li><li><b>Models:</b> Contains discovered models for all clean logs and a random subset of noisy logs (incomplete), using the Alpha, Heuristics, and Inductive miners.</li></ol><p></p>