Abstract: Time series exists widely in various fields. In view of the fact that the existing periodic models are difficult to describe periodic behaviors only activate in partial regions of time series, a novel concept called regional periodic pattern is proposed. After formalizing this regional periodic pattern, a regional periodic pattern mining algorithm based on apriori-like principle is further proposed, which can greatly improve the efficiency of mining under the premise of ensuring the integrity of the mining results. Experiments on a public dataset of network security show that the proposed algorithm can quickly and accurately mine all regional periodic patterns in time series.