One of the criteria for efficient portofilio is that it produces the same level of profit, but with minimum risk.This paper discusses the estimates Value-at-Risk and minimum variance on an investment portfolio.In this case it is assumed that the asset return follows the time series model. Therefore, non-constant meanis estimated using autoregressive moving average (ARMA) models.While non constant volatility is estimated using generalized autoregressive conditionaly heteroscedasticity (GARCH) models. To determine the minimum variance is done using Markowitz's model optimization. Furthermore, Value-at-Risk is calculated based on the values of the mean and minimum variance. The result of return analysis of assets of BBRI, INCI, LPBN, and MPPA, obtained the minimum variance value of 0.011734775, and at the 95% confidence level obtained Value-at-Risk of 0.017873889.The minimum variance and Value-at-Risk are obtained on the vector of the investment weighted composition as x' = (0.092827551, 0.212180907, 0.14631804, 0.548673502). So to get a minimum risk on the investment portfolio consisting of the four assets, the capital allocation must follow the vector of weighted composition produced