A feature extractor of seismic data using genetic algorithms

A novel signal analysis technique is presented and applied for the analysis of seismic data. The main task of the method is the extraction of feature information of a given timeseries. This task is accomplished by the implementation of a specifically designed Genetic Algorithm that is utilised for the evolution of conditional sets. The term conditional set refers to a set of boundary conditions. These boundary conditions must be fullfilled by subsequent samples of the timeseries. The implemented algorithm was applied on seismic data in order to investigate for the existence of conditional sets that appear more frequently than others. It was found that there exist conditional sets with high probability of appearance. Furthermore, it was found that some of the conditional sets are followed by data samples that apper small deviations from their mean value. It is proposed that conditional sets that combine these two properties (high probability of appearance and small deviation of the next data sample) can account for short-term prediction of seismic events.