Utility-Driven Mining of High Utility Episodes
Wensheng Gan, Jerry Chun‐Wei Lin, Han‐Chieh Chao, Philip S. Yu · 2019
Sequence data, e.g., complex event sequence, is more commonly seen than other types of data (e.g., transaction data) in real-world applications. For the mining task from sequence data, several problems have been formulated, such as sequential pattern mining, episode mining, and sequential rule mining. As one of the fundamental problems, episode mining has often been studied. The common wisdom is that discovering frequent episodes is not useful enough. In this paper, we propose an efficient utility mining approach namely UMEpi: Utility Mining of high-utility Episodes from complex event sequence. We propose the concept of remaining utility of episode, and achieve a tighter upper bound, namely episode-weighted utilization (EWU), which will provide better pruning. Thus, the optimized EWU-based pruning strategy can achieve better improvements in mining efficiency. Finally, experiments on two real-life datasets demonstrate that UMEpi can discover the complete high-utility episodes from complex event sequence, while state-of-the-art algorithms fail to return the correct results. Besides, the improved variants of UMEpi outperforms the baseline.