标签传播时间序列聚类的股指期货套期保值策略研究

利用时间序列聚类方法进行股指期货的套期保值，关键要选择合适的聚类方法。本文从新的视角来研究并提高时间序列聚类方法在金融数据分析领域的应用性能，提出一种基于标签传播时间序列聚类的股指期货套期保值模型。该模型以动态时间弯曲为相似性度量方法来构建现货股票网络空间结构，将每只股票看作一个节点，利用标签传播方法将节点划分到不同的簇中，最终实现股票数据聚类。另外，构建最小追踪误差优化模型来确定每支股票在现货组合中的最优权重，从而得到最优组合。实验分别比较新方法和传统聚类方法确定现货组合的追踪误差，结果表明新方法能够提高现货组合的追踪精度，为丰富金融市场投资和管理方式提供新的研究思路。

Research on the stock index futures hedging strategy using label propagation time series clustering

Choosing a suitable clustering method is crucial in using time series clustering in stock index futures hedging. This study aims to investigate and improve the application performance of time series clustering in the financial data analysis field from a new perspective. We propose a model of stock index futures hedging based on label propagation time series clustering. In the model, a network space of spot stock was built using dynamic time warping as similarity measure. Each stock in the network was treated as a node, which would be divided into different clusters using label propagation, and finally, the stock data was clustered successfully. An optimization model for minimizing tracking error was constructed to obtain the optimal weight of each stock in the spot portfolio. Finally, we obtained the optimal spot portfolio. The tracking errors of the portfolio of the proposed method and that of the traditional clustering method were compared by tracking the index in the experiment. The proposed method showed the ability to improve tracking accuracy, providing a new way to enrich the investment and management of financial market.