Stock Investment with Stochastic Programming∗

In this paper, we propose the stochastic programming (SP) model with risk measure conditional value at risk (CVaR) for investing stocks in Taiwan stock market. In each period of investment, 200 scenarios are generated for solving SP, and the CVaR is utilized to manage the risk. The experiment interval starts from 2005/1/1 and ends on 2013/12/31, which has totally 2235 trading periods. The experimental results show that our method is able to earn positive return. We also perform superior predictive ability test to illustrate that our method can avoid the data snooping problem. Our method achieves the best annualized return 13.26%, which is higher than the buy-and-hold annualized return 12.19%. Keywords-Stock; Stochastic programming; Conditional value at risk; Superior predictive ability.