In the real market

(i) in order to prevent market manipulation and other problems caused by short selling

transaction is still subject to some short selling restrictions even in developed markets; (ii) due to the markets relevant regulations and investors’ requirement of risk control

the proportions invested in some assets have certain limits; (iii) investors must pay stamp duty and other transaction costs during transaction. Considering these three aspects

Value-at-Risk (VaR) as risk measure is adopted

and a mean-VaR portfolio model is constructed with limited short selling

proportion of investment limits and transaction cost under two assumptions that the rate of return is normal and non-normal distribution. Firstly

a particle swarm optimization (PSO) algorithm is presented for the model; secondly

numerical experiments are provided by using the test data from A stock market of China; finally

the characteristics of the portfolio efficient frontier and the influences on investors’ decision-making under the limited short selling constraints are discussed.