eBay/KPRN

Combines knowledge graph path reasoning with neural collaborative filtering to generate explainable item recommendations by learning metapath-based embeddings. Built on Python/Lua stack, integrating RNN-based entity reasoning with neural factorization machines to rank candidate items through interpretable graph traversal rather than black-box similarity matching. Designed for training on heterogeneous information networks where recommendation paths can be extracted and validated against user-item interactions.

285 stars. No commits in the last 6 months.