DTW算法在地震时间序列信息挖掘中的应用

Application of DTW Algorithm in Earthquake Time Series Information Mining

采用DTW算法用于实现地震时间序列“异常形态”的回溯性检验与统计,以此提高地震异常信息的深度挖掘效率。随机选取太原地震监测中心站(以下简称太原站)水平摆观测的一段数据作为测试数据,验证该方法的可行性,并利用昔阳地震监测站(以下简称昔阳站)水平摆实际观测数据,自动识别出自观测以来受降雨影响的异常形态。在此基础上,将最近一次数据变化作为模板,进行回溯性检验,共提取出7次类似变化,与之前总结的异常特征一致。经分析,其中4次异常之后对应了太原盆地及其周围M3.6以上地震,运用R值评分对其进行预报效能评估,结果显示该异常形态可通过预报效能检验(R=0.52,R_(0)=0.45),其最佳预测时间为50~90 d内,可作为短期预测指标。

This article uses the DTW algorithm to implement retrospective testing and statistics of "abnormal forms" in earthquake time series,in order to improve the efficiency of deep mining of earthquake anomaly information.A segment of data from the horizontal pendulum observation of Taiyuan Seismic Station was randomly selected as the test data to verify the feasibility of this method.The actual observation data from Xiyang Seismic Station were also used to automatically identify abnormal forms affected by rainfall since self-observation.Based on this,the latest data change was taken as a template for retrospective testing,and a total of 7 similar changes were extracted,consistent with the previously summarized abnormal features.After analysis,4 of the anomalies corresponded to M3.6 or above earthquakes in Taiyuan Basin and its surrounding areas.The R-value score was used to evaluate the predictive performance,and the results showed that the abnormal form could pass the predictive performance test(R=0.52,R_(0)=0.45),with the best prediction time being within 50 to 90 days,making it a short-term predictive indicator.