The Linear Relationship Model with LASSO for Studying Stock Networks.

Area of Science:

• * Quantitative Finance
• * Network Science
• * Computational Economics

Background:

• * Traditional stock market network models primarily focus on pairwise positive correlations.
• * Existing methods overlook the significance of negative correlations and directional influences in market dynamics.

Purpose of the Study:

• * To develop a novel approach for constructing stock relationship networks that includes negative correlations and directional links.
• * To explore the systemic framework of stock markets by analyzing linear relationships using the LASSO (Least Absolute Shrinkage and Selection Operator) method.
• * To investigate network consistency properties using clique blends and balance theory.

Main Methods:

• * Application of the linear relationship model with LASSO to identify both positive and negative correlations.
• * Development of a network model that incorporates statistically significant positive links, negative correlations, and link directions.
• * Utilizing balance theory and clique blends to analyze the consistency of the constructed networks.

Main Results:

• * The proposed method successfully identifies and incorporates negative correlations and directional influences in stock networks.
• * The developed networks demonstrate high consistency with traditional correlation coefficients for both positive and negative relationships.
• * The model provides insights into the direction of influence within the stock market.

Conclusions:

• * The novel network model offers a more comprehensive representation of stock market interdependencies by including negative correlations and influence directions.
• * This approach enhances the understanding of market mechanisms and provides a more robust analytical tool for financial markets.
• * The findings suggest a significant improvement over existing methods that predominantly rely on positive correlations.