PATTERN RECOGNITION OF SEISMICITY BEFORE STRONG EARTHQUAKES

Pattern recognition is a subject that has been developed in the last twenty years and it has already been widely applied in many fields of science. I. M. Gelfand, P. Press and others have applied it to seismic zoning. In this paper, we use the method in earthquake prediction to predict the time of occurrence of an impending strong earthquake.The entire period of time to be investigated is divided into many intervals, each of which is assigned a rating, D for dangerous intervals and N for non-dangerous intervals, according to a definite criterion. The characters of seismicity of medium-size earthquakes before a strong earthquake are put forward by a problem table, then pattern recognition is made in two steps:(1). learning Answers to m questions of P intervals form a mP matrix, binary coded yes or no. By learning. New features to one or combinations of two or three questions, called the characteristics of D and N canbe recognized.(2). Voting The difference of the numbers of D and N characteristics is A. The interval is recognized as being cangerous (D) when △ is equal to or larger than a threshold value, otherwise it is recognized as being non-dangerous (N).The results show that during a definite period of time before a strong earthquake, when the medium-size earthquake activity rises to a certain level and the ratio between the number of earthquakes of a definite magnitude to the number of earthquakes with magnitude a half unit smaller is above normal and also when the number of medium-size earthquakes increases with time, then taking all such characterisitics together into consideration, a strong earthquake would be expected within the next interval of time.In addition, controlled experiments have also been performed, showing that the results of pattern recognition is rather stable.