CVaR model is the alternative method used to measure the risk with a discrete loss, and
asymmetry distribution. This research uses a CVaR model instead of Value-at-Risk (VaR) because
Rockafellar and Uryasev (2000) and Krokhmal et al. (2002) showed the superior results of CVaR
over VaR in measuring the risk and solving the optimization problems by the following advantages. Firstly, CVaR is a continuous and convex function that can be used in linear
programming to find the optimization while VaR can be optimized difficultly due to the discrete
function. Secondly, CVaR is concerned with the risk exceeding than VaR and coherent, especially
the extreme losses that reflect in the tail of distribution. Thirdly, CVaR does not need to assume
normal distribution so we can apply CVaR on asymmetric loss and return distributions.