Identification of Self-Exciting Threshold Autoregressive Model by Using Genetic Algorithm

Nurhidayati, Maulida and Irhamah, Irhamah (2015) Identification of Self-Exciting Threshold Autoregressive Model by Using Genetic Algorithm. In: The 5th Annual Basic Science International Conference, Feb 13th, 2015, Malang.

Preview

Text paper-proceeding-basic-2015-maulida nurhidayati.pdf - Accepted Version Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (1MB) | Preview

Abstract

Self-Exciting Threshold Autoregressive (SETAR) models are models that can be applied to nonlinear time series models. SETAR models are models that partition data into multiple regimes in each regime follow autoregressive model. SETAR model has problem in terms of the identificaton of model. The best model usually obtained by trial and error which allows the best model is not a model of the global optimum. According this problem, in this study used Genetic Algorithm (GA) is a search technique is applied to the target-oriented optimization process to find a global optimum solution. Search techniques performed simultaneously on a number of solutions using three operations: selection, crossover, and mutation. Excellence GA in the optimization process, among others, working on the set of solutions, search based on population of solutions is not only one solution, and use of fitness information to obtain a global optimum solution. Identification model by using GA applied to stock return BMRI. Result from SETAR model using GA are compared with ARIMA model based on AIC. This result shows that SETAR model using GA provides better estimation for daily stock return than ARIMA model, because SETAR-GA produces less AIC then ARIMA.

Actions (login required)

View Item