ARIMA forecasting as a genetic inheritance operator in floating-point genetic algorithms

Abstract

In this paper, a new operator is developed for the floating-point genetic algorithms (FPGAs). The operator records the family tree of chromosomes, searches a convenient time series model on it and forecasts offspring which will possibly be generated by usual genetic algorithm operators in future generations. A software package is developed as an implementation and it is freely available for downloading. The results of a suite of simulation study show that the proposed operator has a statistically significant effect on reaching the global optima of test functions in many dimensions of search spaces. The results of simulation study also show that the developed operator increases the search capabilities of GAs when the number of function parameters increase by means of obtaining the global optimum more precisely.