In this paper, we propose a new prediction analysis model which combines GM (1,1) model from grey system theory and time series ARIMA model from statistics theory. We abbreviate the combined GM (1,1) ARIMA model as ARGM (1,1) model. This combined model takes the advantage of high predictable power of GM (1,1) model and at the same time takes the advantage of the prediction power of ARIMA model. For prediction accuracy improvements, 3-points average model is further applied to ARGM (1,1) model. We call the improved version as 3P-ARGM (1,1) model. In addition, Markov chain model from stochastic process theory to enhance the prediction power of 3P-ARGM (1,1) model. The generated model is called as M3P-ARGM (1,1) model. As an illustrative example, we use the statistical data recording of the business volume of post and telecommunications from 1990 to 2005 for a validation of the effectiveness of the M3P-ARGM (1,1) model.