Which Form Should I Use?
Most data sets on these pages come in one or more of three forms. They hold the same kind of information at different stages of preparation.
| Form | What you get | Pick it when |
|---|---|---|
| Flat event log | One row per event: a case ID, an activity name and a timestamp, often with more columns. | You want to see a process map and start analysing straight away. |
| Object-centric log | Events, the business objects each event touched, and how the objects relate. Published in the OCEL 2.0 standard. | One event touches several things at once (an order, its items, a package) and you do not want to pick a single case ID up front. |
| Raw tables | The tables as the source system or the publisher released them: CSV, Parquet, Excel, or a whole SQLite database. | You want to practise the step before analysis: finding the events in the tables and deciding what a case is. |
Loading each form
| Form | Files you can load | See |
|---|---|---|
| Flat event log | .csv, .xlsx |
Files, OCEL and Event Logs |
| Object-centric log | .json, .jsonocel (OCEL 2.0) |
Files, OCEL and Event Logs |
| Raw tables | A .csv per table, a .zip of them, or one .sqlite file |
Loading Data |
A single file can be up to 2 GB.
Files that need a step first
- XES (
.xes,.xes.gz). Many research event logs are published as XES. Convert the log to CSV with one row per event (case ID, activity, timestamp, then the other attributes) before you load it as a flat event log. - OCEL 2.0 XML or SQLite. Where a publisher offers the same log as JSON, use the JSON file. The SQLite file can also be loaded as raw tables, which is a good way to see how an object-centric log is stored.
- Archives (
.zip,.tar.gz,.7z). Unpack them first. Some hold one log per variant or per year.
Real or simulated
Every data set page says whether the data is real or simulated. Simulated logs are clean and complete, which makes them good for learning. Real logs carry the gaps, duplicates and odd timestamps you will meet in your own data.