Estimation of Vector Autoregressive Model's Parameter Using Genetic Algorithm

One of the multivariate time series models that can be used to estimate process is Vector Autoregressive. However, a problem during the optimal estimating process using the VAR model can result in an inaccurate data. One of the ways to solve the problem is by using optimization during the estimation of the parameter. Genetic algorithm (GA) is one of the optimization methods that can be used to solve the inaccurate data because it creates a global optimum solution. Due to this reason, this research will use Vector Autoregressive for its modeling and be estimating. It will be done by comparing Conditional Least Square (CLS) with GA. The comparing is done by looking at the smallest MAPE value between the results of both estimations of parameter model. This research uses simulation data and the application to the raw data to see the harmony between the information that was received. The application to the raw data uses data of closing stock price from four construction companies that are included in the LQ45 index. The results of the analysis obtained in the simulation data show that GA gives MAPE smaller accuracy on replications data. While in real data result of parameter estimation of GA and CLS is not much different and give an outsample prediction of data approaching an actual data. But if seen from the value of MAPE, GA proved better.