Research Article
Modeling Chaotic Behavior of Chittagong Stock Indices
Table 7
Computational settings for selected forecasting models.
| GA | ANFIS | GARCH |
| (1) Population size—3000, generated at random from the uniform probability distribution. | (1) Input—Lags of cse-all and cse-30 |
cse-all and cse-30: Fitted GARCH (2,1) and GARCH (1,1) Selection Process: BIC. | (2) Fitness function—RMSE of model (2.2.1-2.2.2); Fitness scaling—Linear ranking process. | (2) 2 mfs (type Gaussian) for each of input variables. Thus, 8 if-then fuzzy rules were learned. | | (3) Selection-Roulette wheel. | | | (4) Crossover-fraction: 0.95; Intermediate method. | | | (5) Mutation-fraction: 0.05; Uniform method. | | | (6) Reinsertion: The fitness-based reinsertion method. | | | (7) Termination Criteria—Maximum number of generations, assumed 200. | | |
|
|