Forecasting technique is an important component of decision making because it aims to predict values of data in the future. Many existing forecasting methods only predict single variable data and thus do not consider correlation between variable in a dataset. This paper proposes Vector Autoregressive Moving Average (VARMA) as a forecasting method to predict data with more than one variable. This method combines regression concept i.e. autoregressive (VAR) and moving average method (VMA) for multi-variables data. The first step in VARMA method is testing the stationary of the data. Differencing process is conducted in order to change non stationary data to stationary. The next step is to identify the order of the VARMA model of the stationary data. The parameters are then estimated according to the order and co-integration test is conducted on the variables. The model is tested to assess its validity. If the model is valid then forecasting can be done using the model generic formula. The errors of the forecast are calculated to evaluate the performance of the model. Random values are found in the forecast of VARMA method. However, the error remains within a certain interval. The error interval is below 10% so it can be argued that this model is very accurate in predicting the data.