Predicting stock market movements using network science: an information theoretic approach.

Area of Science:

• Quantitative Finance
• Network Science
• Financial Econometrics

Background:

• Stock markets are complex systems with unpredictable price movements, posing challenges for accurate forecasting.
• Traditional prediction methods often struggle with the inherent complexity and interconnectedness of market dynamics.

Purpose of the Study:

• To develop a new method for forecasting the future movements of the Standard & Poor's 500 Index (S&P 500).
• To leverage time-series complex networks of S&P 500 companies to enhance prediction accuracy.

Main Methods:

• Constructed time-series complex networks using mutual information of 60-min price movements for S&P 500 companies.
• Developed predictive metrics from network strength distributions and centrality measurements.
• Combined network-derived predictors and integrated them into ARIMA models.

Main Results:

• Changes in network strength distributions correlate with future market movements.
• A combined predictor derived from network metrics shows a quadratic relationship with S&P 500 changes, enabling amplitude prediction.
• Integrating network measurements into ARIMA models significantly improved prediction accuracy.

Conclusions:

• The developed network-based approach offers a valuable tool for predicting stock market fluctuations.
• Findings provide actionable insights for financial policy makers and quantitative investors to improve market forecasting and intervention strategies.