h-n-chavda/root-cause-detection

This repository implements causal-graph recovery with the PC algorithm: process data becomes a graph, which can be compared with a reference graph. The documented method separates skeleton discovery from edge orientation, and the documented datasets include two ultra-processed-food cases and Tennessee Eastman. The sampled runner shows loading, graph summarization and optional evaluation; the sampled algorithm excerpt shows skeleton discovery and collider orientation. Meek-rule propagation and the README’s broader run behavior are documented but not visible in the excerpts.