【摘要】 在线投资组合选择是量化金融领域的重要研究问题之一.为建立既有竞争性能理论保证又运行高效的在线投资策略,该文借鉴凸优化技术中的元算法,设计集成反转专家意见的在线投资组合策略.首先,基于股票市场中资产价格的均值反转特征构建专家意见池.其次,运用在线梯度优化方法从理论上分析了策略的后悔度上界,证明了所提出的策略与基准策略相比具有竞争性.最后,从数值上检验了策略在国内外股票市场上的表现.结果表明,该策略在大部分数据集上的累积收益和风险调整收益表现均优于相关的在线策略,甚至超越事后最优的基准策略;具有一定的抗下跌风险能力;且具有线性时间复杂度和运行速度快的优点;并能够在合理的交易费用下保持良好的投资绩效,说明了该策略具有较大的实用性,为投资者提供了一个新方案.更多还原

【Abstract】 Online portfolio selection is one of the fundamental research problems in quantitative finance. To establish an online portfolio strategy that not only guarantees theoretical competitive performance but also runs efficiently, the meta-algorithm in convex optimization technique is used to design a new online portfolio strategy that integrates reversion expert advice. Firstly, a pool of expert advice is constructed based on the mean reversion principle of asset prices in the stock market. Secondly, the upper bound of regret of the proposed strategy is analyzed theoretically by the online gradient optimization method, which proves that the proposed strategy is competitive with the benchmark strategy. Finally, the performance of the proposed strategy is tested numerically in the stock market at home and abroad. The results show that the cumulative wealth and riskadjusted returns of the proposed strategy far exceed those of the relevant online strategies in most datasets, and even exceed those of the optimal benchmark strategies determined in hindsight. Besides, the proposed strategy is able to resist downside risk to some extent. In addition, it has linear time complexity and runs very fast. And the proposed strategy can maintain good investment performance under reasonable transaction costs. It shows that the proposed strategy has practical applicability and provides a new scheme for investors.更多还原