Forecasting of Stock Financial Series Using Multivariate Vector Autoregressive Model

   The ARIMA model (e.g from  'forecast' package for R) is frequently used for forecasting univariate time series.  This model  uses a  previous behavior of only one time series.  Such an approach does not allow for possible relations between different time series. Taking into consideration such  relations, one can give more precise prediction and it can be performed using multivariate vector autoregressive model.

Let us consider an example. Download a historical date of four stock financial series:

then, using multivariate autoregressive model, we can forecast the time series for each stock symbol, taking into account possible relations between these series: