ocelation
ocelation experiments with generating object-centric event data from relational databases by combining graph decomposition and sequence construction.
What it does
- Parses SQL DDL into table/relationship metadata.
- Builds directed graph views of schema dependencies.
- Runs community-detection analyses to partition schemas into meaningful subgraphs.
- Produces sequence and join-generation artifacts for creating merged event-log tables.
How it is built
clustering_utilsandcluster/analysermodules evaluate multiple NetworkX community algorithms.sequenceutilities build spanning-tree-based merge sequences over expanded communities.join/merge.pygenerates SQL join statements and resolves naming collisions.- Notebooks capture exploratory runs, visualization, and prompt-driven LLM experiments.
Tech stack
Python, NetworkX, cdlib, matplotlib, Jupyter notebooks, SQL/DDL parsing.