VynFi/vynfi-supply-chain-ocel

---
license: apache-2.0
task_categories:
- time-series-forecasting
- other
tags:
- synthetic
- financial-data
- vynfi
- process-mining
- ocel
- supply-chain
size_categories:
- 10K<n<100K
configs:
- config_name: events
data_files: "events/train-*.parquet"
- config_name: objects
data_files: "objects/train-*.parquet"
- config_name: anomaly_labels
data_files: "anomaly_labels/train-*.parquet"
- config_name: document_events
data_files: "document_events/train-*.parquet"
---

# VynFi Supply Chain OCEL: 20K Events, 7K Objects

Native OCEL event log from a manufacturing supply chain. 20,010 events across 26 columns, 7,427 objects, 7,961 anomaly labels.

Generated with DataSynth 2.5 OCPM engine. Process types: P2P, O2C, S2C, manufacturing. 3% fraud injection.

## Configs

| Config | Records | Columns |
|--------|---------|---------|
| `events` | 20,010 | 26 |
| `objects` | 7,427 | 9 |
| `anomaly_labels` | 7,961 | 36 |
| `document_events` | 377 | 5 |

## Quick Start

```python
from datasets import load_dataset
events = load_dataset("VynFi/vynfi-supply-chain-ocel", "events", split="train")
objects = load_dataset("VynFi/vynfi-supply-chain-ocel", "objects", split="train")
```

## Limitations

- Event-to-object relationships are stored as stringified arrays.
- The `document_events` config is reconstructed from document flow headers.

## Citation

```bibtex
@dataset{ivertowski_vynfi_ocel_2026,
title = {VynFi Supply Chain OCEL},
author = {Michael Ivertowski},
year = {2026},
url = {https://huggingface.co/datasets/VynFi/vynfi-supply-chain-ocel},
note = {Generated with VynFi (https://vynfi.com)}
}
```

License: Apache 2.0. Entirely synthetic.