Abstract: Portfolio decisions faces a great deal of data of the real security market, which is a complicated combinatorial optimization problem, and is a NP-hard problem, which is difficult to be solved by traditional algorithm. Cultural algorithm and particle swarm optimization are emerging bionic intelligence algorithms. This paper introduces a new dynamic particle swarm optimization based on cultural algorithm for solving the mean-VaR model, and uses the penalty function approach to the inequality constraints in the model. An empirical analysis is done by sixteen securities chosen from Shanghai and Shenzhen security markets as the alternative securities. The numerical results show that the problem of portfolio optimization can be solved more reasonably and efficiently by this algorithm.

Key words: particle swarm optimization, cultural algorithm, portfolio, mean-Value at Risk（VaR）

摘要： 投资组合决策面临现实证券市场中的大量数据，是一个复杂的组合优化问题，属于NP难问题，传统的算法难以有效求解。文化算法和粒子群算法是新近出现的两种仿生智能算法，将新提出的动态文化粒子群算法用于求解均值-VaR模型，用罚函数方法处理模型中的不等式约束，选取沪市和深市的十六支股票作为备选股票进行实证分析，数值结果表明该算法可以高效、合理地解决投资组合优化问题。

关键词: 粒子群优化算法, 文化算法, 投资组合, 均值-风险价值（VaR）