基于遗传神经网络的指数跟踪优化方法

针对使用完全复制法进行指数跟踪的缺点和仅以跟踪误差作为指数跟踪目标的不足,以跟踪误差最小化和超额收益最大化两者的权衡作为指数跟踪的目标函数,综合考虑实际中的交易成本、现金、卖空限制等约束,建立指数跟踪优化模型,并采用二进制和实数值混合编码的遗传BP网络对指数跟踪管理中的资金进行优化配置.该算法能同时优化网络结构和权值矢量,并结合遗传算子和Solis Wets算子生成后代使遗传搜索空间的群体多样性更好,加快了遗传算法的收敛速度,提高了连接权系数的优化精度.跟踪深证100指数的实证结果表明:应用遗传神经网络算法进行指数跟踪的效果明显优于完全复制法,并且实现了目标指数的动态跟踪.

Index tracking optimization method based on genetic neural network

For the disadvantages of full replication approach and only adopting tracking error as the object of index tracking,an index tracking optimal model is constructed in which a tradeoff between minimizing tracking error and maximizing excess return is treated as objective function,and practical constraints such as total transaction cost limit,cash flow constraint and short sale restriction are considered.Moreover,an improved genetic neural network using binary and reai-vaiued combination coding method is prese...