A Robustness of Vector Auto-Regressive Models Performance On Outliers Detection Using Genetic Algorithm

In this paper, a robustness of Vector Auto-Regressive (VAR) models parameter will be estimated by using genetic algorithm (GA) on outliers detection. Least Square (LS) estimator has been adopted in GA term to estimate a robust parameters of the VAR models were represented by chromosomes in GA's term. The chromosomes are represented by autoregressive coefficient matrix that changes on each different lag time and requaired a checking models to guarantee a stasionerity of the models when the outliers presented in a multivariate time series data. Based on these simulation, the models performance which compared via Mean Square Error (MSE), Akaike Information Criteria (AIC) and its time consumptions. These results are not only remain model performance on ouliers detection, but also improving a robustness of the parameter models