核心概念
The author proposes the COPP-Miner algorithm for top-k contrast pattern mining to improve time series classification by discovering patterns with significant differences between classes efficiently.
要約
The content introduces the COPP-Miner algorithm for top-k contrast pattern mining in time series classification. It addresses the limitations of frequent patterns by focusing on contrast patterns. The algorithm consists of extreme point extraction, forward mining, reverse mining, support rate calculation, and pruning strategies. Experimental results validate the efficiency of the proposed approach.
統計
COPP-Miner is composed of three parts: extreme point extraction to reduce time series length, forward mining, and reverse mining.
Experimental results show that top-k COPPs can be used as features for better classification performance.