Abstract
AbstractIn this paper, the application of genetic algorithm (GA) in linear and non-linear dynamic modeling, as well as the development of an alternative GA-based model structure selection algorithm is presented. As a benchmark for the proposed algorithm, orthogonal least square (OLS), a gradient descent method, is utilized. To reduce problems of premature convergence in simple GA (SGA), a model structure selection based on modified genetic algorithm (MGA) is proposed. The effect of different combinations of MGA operators on the performance of the developed model was studied, and the effectiveness and shortcomings of MGA were highlighted. Results were compared between SGA, MGA, and OLS. From the results obtained, it was observed that with similar number of dynamic terms, MGA performs better than SGA, in terms of exploring potential solution, meanwhile outperformed the OLS algorithm in terms of selected number of terms and predictive accuracy. In addition, for fine-tuning, the use of local search of memetic algorithm (MA) with MGA is also proposed and investigated. From the results obtained, it can be observed that MA can produce an adequate which satisfy the model validation tests with significant advantages over OLS, SGA, and MGA methods.KeywordsGenetic algorithmOrthogonal least squareModified genetic algorithmSimple genetic algorithmArtificial intelligence
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
