Synthetic Relational Databases and Reference OCELs
Description
This dataset provides six deterministic, fully synthetic relational source-system databases for benchmarking the extraction of object-centric event logs from relational data. Each database is paired with a canonical reference OCEL in all five OCEL 2.1 formats.
Synthetic relational databases and reference OCELs
The source databases are not event logs. Business milestones are distributed across domain-specific tables alongside technical state, duplicates, corrections, retries, historical rows, projections, and redundant evidence. The reference OCELs demonstrate how that evidence can be filtered and combined into canonical business events with normalized UTC timestamps, stable source provenance, and relations to the relevant business objects.
All data is synthetic and was generated deterministically with seed 20260820.
Scenarios
| Scenario | Domain | Primary / secondary case notion | Events | Objects |
|---|---|---|---|---|
| ForgeFlow | Engineer-to-order manufacturing | SalesOrder / Shipment | 97,884 | 59,036 |
| TrialVersion | Clinical-trial execution | Participant / Sample | 74,661 | 33,858 |
| ProcureChange | Industrial procure-to-pay | PurchaseRequisition / Invoice | 100,504 | 46,353 |
| ClaimStream | Insurance claims | Claim / Exposure | 99,715 | 40,610 |
| PermitFlow | Building-permit approval | PermitApplication / Inspection | 118,084 | 32,783 |
| BatteryVault | Circular battery lifecycle | BatteryPack / ServiceOrder | 88,346 | 25,779 |
The scenarios represent several relational persistence patterns, including normalized OLTP, in-table version histories, SAP-style change documents, append-only event stores, workflow-engine histories, and a bitemporal Data Vault. Every scenario contains 3,000 primary instances.
Files
For each scenario, the dataset includes:
- A relational source-system database in SQLite format (
<name>.sqlite). - A gzip-compressed XML OCEL (
<name>.xmlocel.gz). - A gzip-compressed JSON OCEL (
<name>.jsonocel.gz). - An OCEL SQLite database (
<name>.ocel.sqlite). - A compact OCEL 2.1 CSV file (
<name>.ocel.csv). - An OCEL 2.1 bundled ZIP using typed Parquet tables (
<name>.ocel.zip).
The five OCEL files for a scenario are equivalent alternative serializations, not shards. They include qualified event-to-object and object-to-object relations, temporal object-attribute changes, activity-specific event attributes, resource roles and calendars, exception paths, and source table and record provenance.
The archive also provides validation.json with machine-readable validation
results and SHA256SUMS for integrity verification.
Potential Uses
- Benchmarking relational-to-OCEL extraction methods.
- Evaluating event identification, filtering, and evidence-fusion strategies.
- Testing support for the five OCEL 2.1 serialization formats.
- Comparing extracted logs with canonical reference OCELs.
- Teaching object-centric event extraction from complex relational systems.