时间序列的趋势转折点提取算法及应用研究

Research and Application of Trend Turning Point Extraction Algorithm for Time Series Data

针对时间序列趋势转折点的提取,提出一种基于矢量转角和极值点相结合的改进算法,并应用到地震观测资料的分析中。结合地震对应规则,对山西断陷带北部小磨流动水准XM1-XM2测线的地震预测效能进行评估。结果表明:1)该算法可用于时间序列趋势转折点的自动提取,提高了人工判别的工作效率,实现了转折点的定量化识别;2)从数据统计的角度考虑,该测线与200 km范围内M4.5以上地震有较好的对应关系,预测效能评估结果R=0.63(R0=0.43);3)2000年以后该测线的地震虚报率较高,分析认为可能与远场强震引起的应力变化有关。

In order to extract the trend turning point of time series,we propose an improved algorithm based on the combination of vector turning angles and extreme points applied to the analysis of seismic observation data.Combined with the corresponding rules of earthquakes,we evaluate the seismic prediction effectiveness of the XM1-XM2 line of Xiaomo mobile leveling,located in the northern Shanxi fault depression.The results show that:1)The algorithm can be used to automatically extract the trend turning points of the time series,improving the efficiency of manual discrimination,and achieving the quantitative identification of turning points;2)From the perspective of data statistics,the XM1-XM2 line has a good corresponding relationship with M>4.5 earthquakes within 200 km.The evaluation result of prediction effectiveness is R=0.63(R0=0.43);3)The seismic inflating rate of this line after 2000 is higher,probably related to the stress change caused by the far field effect from strong earthquake.