论文部分内容阅读
本文将目前流行的规则化方法加入到传统指数追踪模型中,得到若干种稀疏而且稳定的资产组合,用于复制指数的收益率,并构建样本内外预测、模型一致性、资产组合稀疏性和BIC准则进行模型效果评价。基于对上证综指、沪深300指数和中证500指数的实证发现:图结构约束可以提升模型的样本外预测能力、模型一致性和资产组合稀疏性;ITM-adaL1在资产组合稀疏性上表现远好于其他模型;结合三种指数追踪,含有自适应L1罚函数以及图结构约束的指数追踪模型总体表现优于其他模型。本文的研究方法和结果对指数型基金管理公司、个人和投资机构者有较为重要的实际意义。
In this paper, the current popular regularization method is added to the traditional index tracking model to obtain a number of sparse and stable portfolio of assets, which is used to replicate the return rate of the index and to build the sample internal and external forecasting, model consistency, portfolio sparsity and BIC Criteria for model effect evaluation. Based on the empirical findings of the Shanghai Composite Index, the CSI 300 Index and the CSI 500 Index, graph structure constraints can improve the model’s out-of-sample forecasting ability, model consistency and portfolio sparsity; ITM-adaL1 represents the portfolio sparsity Compared with other models, the index tracking model with three kinds of exponential tracking and adaptive L1 penalty function and the graph structure constraint is better than other models. The research methods and results of this article have more practical significance for index fund managers, individuals and investment institutions.