Application of genetic algorithm to pattern extraction

The area of interest for this paper covers pattern recognition method, which can find and classify all useful relations between data entries in the time series. Genetic algorithm has been deployed to prepare and govern a set of independent patterns. For each pattern additional quality value has been added. This value corresponds to the level of certainty and is introduced in the work. Practical application of this solution consists of data fitting and prediction. Analyzed data can be non continuous and incomplete. In uncertain cases algorithm presents either no response at all or more than one answer to processed data. Architecture of the system offers possibility to interleave learning phase with use. Genetic algorithm applied in the method facilitates niche techniques as well as crowd factor and specialized population selection methods. Early testing results, which include prediction and fitting of simple time series with up to 50 percent of missing data, are presented at the end of the paper.