This paper proposes a comprehensive framework for efficient GPAR mining, integrating pattern reduction techniques, anti-monotonicity-based pruning, and hybrid sampling strategies. The novel algorithm GraphPAR leverages auxiliary graph structures and maximal independent set-based support counting to achieve substantial performance improvements over existing methods.
Key findings
GraphPAR extends traditional association rule mining to graph-structured data.
The algorithm addresses memory overhead, redundant pattern generation, and scalability issues on billion-edge graphs.
GraphPAR achieves theoretical guarantees on confidence bounds and delivers up to 10.58x speedup compared to state-of-the-art baselines.
Limitations & open questions
The paper does not discuss the limitations of GraphPAR in detail.