Application of genetic and fuzzy modelling in time series analysis

Abstract

Researchers have proposed several change point detection and testing methods. However, in real cases, it has been shown that the structure of a time series changes gradually, i.e. the change points illustrate a sense of fuzziness. This research is based on the concept of a time series model and on fuzzy theory. It combines the concept of genetic models with other leading models. We use time series statistical models as chromosomes in the process of genetic evolution, and also use the membership functions of selected models as a performance index of the chromosomes. Change point analysis could be helpful in fitting different models to different data regimes. These models could then be used for forecasting future time series data using a genetic algorithm approach instead of using only the last model. Also, different models at different time periods could give some insight regarding an economic interpretation of the data during that regime.