The AS index is a new risk measure put forward recently by Aumann and Serrano who are inspiredby the theory of choice under uncertainty． It has many advantages over other risk measures andattracts many scholars． In this paper，we consider an asset allocation problem with the Mean-AS model under normal distribution and general distribution assumption，respectively． In the former case，we obtain an analytical expression of portfolio frontier and thoroughly discuss the characteristics of portfolio frontier． In the latter case，we embody the AS moment estimator into the Mean-AS portfolio optimization model and implement risk estimation and portfolio optimization simultaneously． Under very mild conditions，we prove that the Mean-AS model is a convex optimization problem and an iterative algorithm can be designed to obtain its numerical solution． Monte Carlo simulation results show that the Mean-AS model and our algorithm are accurate and effective． Finally， an empirical case of stock portfolio in Chinese A-stock market is illustrated．