An Investigation into the Effect of Number of Model Parameters on Performance in Type-1 and Type-2 Fuzzy Logic Systems
Salang Musikasuwan, Turhan Ozen, Jonathan M. Garibaldi · 2004
An investigation was carried out in which the performances of type-1 and type-2 fuzzy logic systems (FLSs) with varying number of tunable parameters were compared in their ability to predict the Mackey-Glass time series with various levels of added noise. Each of the FLSs were tuned to achieve the best possible performance using a standardised gradient descent procedure. These experiments were repeated a number of times in order to establish the mean performance of each FLS. The results show that the best performance was achieved with a type-1 FLS, albeit featuring a high number of tunable parameters. A type-2 FLS with far fewer parameters achieved performance very close to the best. Keywords: Type-1 fuzzy logic systems, type-2 fuzzy logic systems, time-series forecasting. 1