KGBoost: A classification-based knowledge base completion method with negative sampling

Knowledge base completion is formulated as a binary classification problem in this work, where an XGBoost classifier trained for each relation using relevant links knowledge graphs (KGs). The new method, named KGBoost, adopts modularized design and attempts to find hard negative samples so train powerful missing link prediction. We conduct experiments on multiple benchmark datasets demonstrate that KGBoost outperforms state-of-the-art methods across most datasets. Furthermore, compared with models by end-to-end optimization, works well under the low-dimensional setting allow smaller model size.