• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International Conference on Parallel and Distributed Processing Techniques and Applications(PDPTA'03) v.2; 20030623-20030626; Las Vegas,NV; US >Multi-objective Multi-stage Portfolio Optimization with Parallel GA
【24h】

Multi-objective Multi-stage Portfolio Optimization with Parallel GA

机译：并行遗传算法的多目标多阶段投资组合优化

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Here -we present a parallel solver for obtaining Pareto-optimal solutions for portfolio construction and rebalancing. Since the search space is very extensive, parallel GA is deployed as our self-learning portfolio optimizer. Firstly GA performs screening of financial instruments based on their recent performance with respect to return, risk and maximum drawdown. Secondly GA is applied to the trading model to optimally rebalance the portfolio by buying low and selling high. The first part uses a multi-objective fitness Junction on the portfolio whereas the second part uses a multi-objective fitness Junction for trading strategies for each of the constituent components of the portfolio. The multi-objective junctions aim at maximizing return per unit of risk alongwith minimization of maximum drawdown. For GA we basically use an Island Model with a migration operator, which allows the fittest sub-population on different nodes to mingle with each other so as to avoid premature convergence. Due to the complexity of the multi-objective function analysis and the size of the search space, we use a parallel implementation of GA on a cluster of quad-processor machines. Due to the parallelisation techniques adopted for workload distribution and inter-node communications, the implementation yields high order of speed-up over sequential implementation and yields better quality solutions in reasonable time.
机译：在这里，我们提出了一个并行求解器，用于获得帕累托最优解，以进行投资组合构建和再平衡。由于搜索空间非常广泛，因此并行GA被部署为我们的自学产品组合优化器。首先，通用汽车根据金融工具在回报，风险和最大亏损方面的近期表现进行筛选。其次，将GA应用于交易模型，以通过低买高卖来最佳地平衡投资组合。第一部分使用投资组合上的多目标适应性结点，而第二部分使用投资组合中各个组成部分的交易策略使用多目标适应性结点。多目标联结旨在最大程度地降低单位风险收益，同时最大程度地减少最大亏损。对于GA，我们基本上使用带有迁移运算符的Island模型，该模型允许不同节点上的适体子种群相互混合，以避免过早收敛。由于多目标函数分析的复杂性和搜索空间的大小，我们在四处理器计算机集群上使用GA的并行实现。由于工作负载分配和节点间通信采用了并行化技术，因此与顺序实施相比，该实施产生了很高的加速比，并且在合理的时间内产生了质量更高的解决方案。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号