Predictability of Complex Systems

Abstraction: Review of theoretical limits and methods for predicting complex systems

Key points:

• Focuses on predictability limits — theoretical upper bounds on prediction accuracy — rather than just prediction algorithms
• Covers three core domains: predictability of time series, predictability of network structures, and predictability of dynamical processes
• Predictability itself serves as a feature for characterizing complex systems and as a benchmark for evaluating prediction algorithms
• Integrates both data-driven approaches (from the data science era) and mechanistic/analytical perspectives
• Accurate predictability estimation reveals the gap between current algorithm performance and theoretical limits, guiding algorithmic improvement
• Applications across natural sciences, social sciences, and engineering; includes open challenges for future research

Connections: Complex Systems · Predictability · Time Series Forecasting · Network Dynamics · Dynamical Systems

Source: https://arxiv.org/abs/2510.16312