OPR-Miner: Order-preserving rule mining for time series

Discovering frequent trends in time series is a critical task data mining. Recently, order-preserving matching was proposed to find all occurrences of pattern series, where the relative order (regarded as trend) and an occurrence sub-time whose coincides with pattern. Inspired by matching, existing (OPP) mining algorithm employs calculate support, which leads low efficiency. To address this deficiency, paper proposes called efficient OPP miner (EFO-Miner) OPPs. EFO-Miner composed four parts: fusion strategy generate candidate patterns, process for results sub-patterns support super-patterns, screening dynamically reduce size prefix suffix arrays, pruning further prune patterns. Moreover, explores rule (OPR) OPR-Miner discover strong rules from OPPs using EFO-Miner. Experimental verify that gives better performance than other competitive algorithms. More importantly, clustering classification experiments validate achieves good performance.