Dynamical analysis of financial stocks network: Improving forecasting using network properties.

Area of Science:

• Quantitative Finance
• Network Science
• Financial Econometrics

Background:

• Stock market dynamics are complex and influenced by intricate interactions between individual stocks.
• Traditional models often overlook the network structure inherent in stock correlations.
• Understanding these network properties is crucial for accurate market prediction.

Purpose of the Study:

• To apply network analysis to stock return correlations to understand market dynamics.
• To identify network-based variables that correlate with future stock returns.
• To enhance stock return forecasting models using network properties.

Main Methods:

• Network analysis of stock return correlations.
• Identification of dynamical network properties.
• Correlation analysis between network variables and S&P 500 stock returns.
• Forecasting stock returns using network-derived input variables.

Main Results:

• Identified meaningful network variables capturing stock interactions and market structure.
• Achieved a 21% improvement in R2 score for yearly stock return prediction.
• Achieved a 3% improvement in R2 score for 2-day stock return prediction.
• Demonstrated the predictive power of network-based variables over baseline models.

Conclusions:

• Integrating network-based variables significantly enhances stock return prediction accuracy.
• Network analysis provides deeper insights into complex financial market dynamics.
• This approach offers valuable tools for investors and researchers in financial modeling and prediction.