Abstract:

In order to solve the problem of low detection accuracy caused by the lack of abnormal data in the existing precursor anomaly detection methods, a detection method based on negative selection is proposed. Firstly, self set and nonself set in seismic data are defined. Secondly, the randomly selected immature detector is matched with self set to generate a maturity detector with variable radius to cover nonself space. Then, the data to be detected are matched with the detector to determine whether the detection result is obtained in nonself space. Finally, compared with the existing seismic anomaly detection methods, BP neural network and support vector machine. The experimental results show that the negative selection is more effective for the anomaly detection of seismic precursor observation data.

Key words: negative selection, anomaly detection, earthquake precursor observation data, computer immune system

摘要：

针对现有前兆异常检测方法因异常数据较少导致检测准确率偏低的问题，提出一种基于反向选择的检测方法。定义地震数据中的self集与nonself集；将随机选取的未成熟检测器与self集进行匹配，生成半径可变的成熟检测器，覆盖nonself空间；将待检测数据与检测器匹配，通过判断是否在nonself空间得到检测结果；与现有地震异常检测方法BP神经网络、支持向量机进行对比，实验结果表明反向选择用于地震前兆观测数据异常检测有更好的效果。

关键词: 反向选择, 异常检测, 地震前兆观测数据, 计算机免疫系统