金融高频时间序列的MODWT波动分析

Financial High Frequency Time Sequence MODWT Fluctuation Analysis

与经典小波变换相比，利用最大交叠小波变换（MODWT）对非平稳时间序列进行分解时，由于没有下采样的过程，因此可以最大限度地减少数据信息的遗失。该文通过对股指期货主力合约一天中的采样数据连行研究。发现MODWT可以有效地对序列中的波动与趋势进行分解。此外文章中还发现，如果分解层数足够多，那么大部分的趋势信息则被波动信息所覆盖。因此总结出用小波对零均值数据进行滤波时，要适当选择分解的层数。

Compared with the classical DWT, the maximum overlap wavelet transform is non-decimated, so when decompose time series, it can reserve the maximum data information. This paperresearches the daily trading amount data of stock index future top contact and found that MODWT can effectively decompose the fluctuations and trends. There is another finding that if.the level of decomposition is enough, the trend information can be covered by the fluctuations. So when. wavelet is used as the filter for the zero-mean data, the level should be selected appropriately.