Optimisation of transfer function models of multi input single output systems using genetic algorithms

Abstract

Initial transfer function estimates usually require optimisation. In order to obtain a more efficient transfer function model, a manual estimation method is often employed. This manual estimation method, even though it results to better estimates of the transfer function model, it still does not give an optimum outcome. In order to obtain an optimum transfer function estimate, open source software based on genetic algorithm was developed. The software was developed with Visual Basic programming language. In order to test the software, a transfer function model was developed from data obtained from industry. The forecast obtained from the transfer function model has a MAPE of 99.82%. After optimisation using the developed software the MAPE of the new forecast was 99.84%. This shows a marginal improvement in forecast accuracy.