论文部分内容阅读
本文用五种有序集群方法研究了华北地区和宁夏地区强震的孕震过程。我们得到:五种方法均能模拟地震活动交替的高潮、低潮现象;Fisher 方法和有序点群分析方法(采用离差平方和增量作为类间距离时)的效果均较好,表明在各类中诸时间段的结构主要是团状分布而不是链状分布,因此选择预报方法时考虑地震活动的这一特点是重要的;对分移动窗口法和非参数拟合优度法的结果,表明在相对平静期和显著活动期之间确实存在着客观的谷值;图论方法则有助于进一步分出显著活动期中易发生强震的较小峰值。五种方法中以 Fisher 方法的误识率(?)最低,而非参数拟合优度法则能帮助确定较为合理的分类。此外对资料预先平滑也可改善结果。
In this paper, five kinds of ordered cluster methods are used to study the seismogenic processes of strong earthquakes in North China and Ningxia. We obtain that all five methods can simulate alternating climaxes and ebb phenomena of seismicity. The Fisher method and ordered point-group analysis method (using the sum of squared deviations of squared distances as the distance between classes) Therefore, it is important to consider the characteristics of seismicity in the selection of forecasting methods. For the results of sub-moving window method and non-parametric fitting goodness-of-fit method, Which indicates that there is indeed an objective valley between the relatively quiet period and the significant period of activity. The graph theory method helps to further separate the smaller peaks of strong earthquakes that are likely to occur during the significant period of activity. Among the five methods, the Fisher method has the lowest false-positive rate (?), While the non-parametric goodness-of-fit method can help determine the more reasonable classification. In addition to the data pre-smoothed can also improve the results.