FS方法及其在综合多项震兆中的应用

FS METHOD AND ITS APPLICATION IN EARTHQUAKE REGIONS WITH MULTIPLE INDICATIONS

本文提出了一种模式识别方法——FS方法.该方法的特征取为某些论域上的模糊子集.先以每三个原始特征组成子分类器的特征集,“训练”这些分类器,然后对子分类器进行筛选,最后用筛选出的子分类器的“软”分类结果的加权平均作为判别函数,形成总的分类决策.将FS方法应用于综合多项震兆,并进行了一系列控制试验,还与地震学中几种常用的方法进行了对比试验.结果表明,FS方法是稳定和有效的.

In this paper a pattern recognition method, namely FS methol, is proposed. The features In FS method are fuzzy sets on some universes. Take every set consisting of three original features as one feature set, training a classifier called sub-classifier. Select a part of sub-classifer from all of them. Take weighted mean of the 'soft' classification of the sub-classifier selected as discrimination function, formulating ultimate recognition rule. This method is used to synthesize multi-precursor prior to strong earthquakes. A series of control tests are taken and the experiment comparing with other several common methods is done. It is show by the results that FS method is stable and efficient.