均值-CVaR投资组合模型及改进的蝙蝠算法求解

An improved bat Algorithm for Solving Mean-CVaR Portfolio Model

基于投资者是风险厌恶型和风险资产价格路径服从跳扩散过程的假设,采用条件风险价值来度量组合风险,建立均值-CVaR投资组合优化模型.为快速有效求解模型,将基于模型的交叉熵随机优化方法嵌入到基于群体的蝙蝠仿生算法中,构建一种改进的蝙蝠算法,该算法既充分发挥交叉熵方法的随机性、自适应性和鲁棒性,又有效抑制蝙蝠算法的早熟收敛现象.借助Monte Carlo模拟情景生成得到价格路径,进而采用所建算法实现模型求解,并与遗传算法和线性规划方法进行比较.实验结果表明,新算法在求解有效性和实用性方面表现更好,取得更为满意的结果.

In this paper we assume that investors are risk averse and risk assets are driven by jump-diffusion processes. And a mean-CVaR portfolio selection model was proposed in which Conditional Value-at-Risk was used as a method to measure the portfolio risk An improved bat algorithm was proposed by which the model-based cross-entropy stochastic optimization algorithm is embedded population-based bat algorithm from bionics for solving the proposed portfolio model. The improved algorithm fully absorbs the stochastic, adaptability and robustness of cross-entropy, and adaptively avoids the stagnancy of population and increase the speed of convergence. Monte Carlo simulation was employed for generating scenario paths based on jump- diffusion model. The improved algorithm was compared with LP, and GA in optimization efficiency. The empirical results demonstrate that the proposed algorithm can achieve satisfactory results and perform well in solving the mean-CVaR portfolio selection model.