ESTIMATION OF LOGISTIC SMOOTH TRANSITION AUTOREGRESSIVE MODEL USING GENETIC ALGORITHM

Time series data not only create linear model but also nonlinear model, especially in the economic. One of nonlinear model in time series is smooth transition autoregressive (STAR). STAR model is the development of Self-Exciting Autoregressive (SETAR) model with two regimes. STAR smoothen the SETAR two regimes with a transition function and one of that is logistic function which then form LSTAR models (logistic smooth transition autoregressive). After, determining order p and delay d, estimation of π1, π2, γ and c using genetic algotihm. By GA (genetic algorithm), that is a technique to search for solutions with various combinations of genes in a chromosome in order to obtain a global optimum solution, . Evaluation of chromosomes carried by the fitness function through procedures such as selection, crossover and mutation. Application of genetic algorithm (GA) will be conducted to identify the data model of LSTAR on stock return are included in the LQ 45 index that show LSTAR with GA better than AR model. Key words: genetic algorithm, nonlinear, stock return, LSTAR