基于自相似的金融时间序列波动聚集性研究

The Research of Clustering of Volatility of Financial Time Series Based on Self- Similarity

自相似与波动聚集性是金融时间序列的两个重要特征,文章将这两个特征结合,提出了一种基于自相似的波动聚集模型。基于该模型提出了一种基于拟合优度与趋势变动的联机时间序列分割算法,算法能够根据波动的自相似特征将序列分割为多个子序列,从而用于研究在不同时段金融时间序列波动的自相似性。对实际数据的实验结果表明,文章所提出的模型和分割算法是有效的。

Self-similarity and clustering of volatility are two important characters of financial time series.This paper combines these two characters,and then brings forward a model of clustering of volatility based on self-similarity.Based on the model,an online segmentation algorithm based on the quality of fit and the change of trends for time series is proposed.The algorithm can segment series by the character of clustering of volatillty,so it can be used to study the self-similarity of financial time series during different time.The experiments conducted on real-world datasets show the model and algorithm are practical.