Research On Time Series Correlation Based On Sequence Alignment Model

The real-time fluctuations of financial asset prices reflect the current changes in financial market risks.It is financial mathematics to find the law of asset price changes in different financial markets,especially from the correlation law of time series of different financial markets.Research hot issues.In order to better meet the needs of the market,this article conducts an in-depth study on identifying hidden patterns between time series.Based on the theory of sequence alignment model(SA)and generalized autoregressive conditional heteroscedasticity model(GARCH),this paper proposes a new sequence alignment model(GARCH-SA)for determining the correlation of time series.The construction process of the new model is to first explore the common occurrence probability of symbol pairs in the sequence to be compared,redefine the Blosum matrix to initially explore the hidden patterns existing between time series,and then use this newly defined Blosum matrix is used for sequence comparison to construct a new SA model and combined with the data features output by the GARCH model to obtain the final GARCH-SA model.Empirical research shows that the new model can effectively improve the reliability of information in symbolized time series,and can capture the different leading-lag characteristics of different financial markets and other pattern information.At the same time,it can be used for financial time series data without obvious periodicity.The problem of missing data in different situations has better robustness and can obtain better matching results.In addition,this paper uses the sequence matching method output by the GARCH-SA model to combine with the multidimensional scaling(MDS)method,and proposes a visualization method based on the correlation of the sequence alignment model(SA-MDS).Compare the new SA-MDS method with the traditional multi-dimensional scaling method of Euclidean distance dissimilarity and the multi-dimensional scaling method of Spearman's correlation coefficient to study the correlation of financial time series and analyze the different similarities In the case of measurement,their category is displayed.The empirical study found that,compared with the traditional multi-dimensional scaling method of Euclidean distance dissimilarity and the multi-dimensional scaling method of Spearman's correlation coefficient,the newly proposed SA-MDS method has an effect on the correlation between data of different sectors in the financial market.It gives a more intuitive result,and can further classify hidden modes such as lead and lag between different time series.