The Linear Relationship Model with LASSO for Studying Stock Networks

Entropy

Hongjiong Tian¹  Boyao Wu²  Muzi Chen³  Tianhai Tian⁴ 

[1] Department of Mathematics, Shanghai Normal University, Shanghai 200234, China;School of Banking and Finance, University of International Business and Economics, Beijing 100029, China;School of Management Science and Engineering, Central University of Finance and Economics, Beijing 102206, China;School of Mathematics, Monash University, Clayton, VIC 3800, Australia;

关键词: stock relationship network;    linear relationship model;    LASSO;    negative correlation;   

DOI  :  10.3390/e24060808

来源: DOAJ

【 摘 要 】

The correlation-based network is a powerful tool to reveal the influential mechanisms and relations in stock markets. However, current methods for developing network models are dominantly based on the pairwise relationship of positive correlations. This work proposes a new approach for developing stock relationship networks by using the linear relationship model with LASSO to explore negative correlations under a systemic framework. The developed model not only preserves positive links with statistical significance but also includes link directions and negative correlations. We also introduce blends cliques with the balance theory to investigate the consistency properties of the developed networks. The ASX 200 stock data with 194 stocks are applied to evaluate the effectiveness of our proposed method. Results suggest that the developed networks not only are highly consistent with the correlation coefficient in terms of positive or negative correlations but also provide influence directions in stock markets.

【 授权许可】

Unknown