Financial market predictability with tensor decomposition and links forecast.

Area of Science:

• Quantitative Finance
• Complex Systems Analysis
• Financial Econometrics

Background:

• Financial markets exhibit complex interdependencies captured by correlation networks.
• Market turbulence and crises cause regime shifts, increasing stock correlations and network interconnectedness.
• Understanding these shifts is crucial for risk management and portfolio optimization.

Purpose of the Study:

• To predict abrupt changes in financial markets by forecasting future stock price dynamics.
• To identify spatial signals indicative of market turbulence and critical transitions.
• To enhance portfolio optimization strategies by minimizing interconnectedness risk.

Main Methods:

• Utilizing complex network perspective to analyze stock price time series.
• Applying tensor decomposition techniques to extract interdependencies and relationships.
• Developing a portfolio maximization application based on predicted stock price distances.

Main Results:

• Empirical evidence of spatial signals, specifically increasing spatial correlation, near critical transitions.
• Demonstrated ability of the methodology to forecast future stock prices using spatial signals.
• Optimization approach minimizing interconnectedness risk yields superior investment plans compared to simpler strategies.

Conclusions:

• Increasing spatial correlation serves as a predictive signal for market turbulence.
• Tensor decomposition-based portfolio optimization effectively mitigates interconnectedness risk.
• Minimizing asset interconnectedness in portfolios enhances resilience against market shocks.