Research on Identifying Causal Relationships in High-dimensional Sequences Based on Elastic Network Regularization
The Elastic Net regularization method combines the advantages of both L1 and L2 regularization,effectively addressing issues such as collinearity,overfitting,and information redundancy in the analysis of causality in high-dimensional time series.It efficiently uncovers the primary correlations among high-dimensional sequences.First,the methodology for identifying causal relationships in high-dimensional sequences was elaborated based on the VAR(Vector Autoregression)model and the long-term Granger causality network.Second,how Elastic Net regularization addresses the shortcomings of causal modeling for high-dimensional sequences by incorporating the key features of both L1 and L2 regularization was discussed.Lastly,the application of Elastic Net regularization in identifying causal relationships among high-dimensional sequences was explored using daily volatility sequences of multi-dimensional commodity futures prices,and compared with other regularization methods.The results demonstrate that the Elastic Net regularization method avoids issues of information redundancy and"excessive sparsity,"identifying and refining the primary influential nodes and transmission relationships within the causal network of commodity price fluctuations,thereby uncovering the major causal associations within the sequence relationship network.
万方数据
维普
