SPADE4: Sparsity and Delay Embedding Based Forecasting of Epidemics.

Area of Science:

• Epidemiology
• Computational Biology
• Data Science

Background:

• Predicting infectious disease evolution is difficult due to scarce and incomplete data.
• Compartmental models are common but may oversimplify complex disease dynamics and human interactions.

Purpose of the Study:

• To introduce Sparsity and Delay Embedding based Forecasting (SPADE4) for improved epidemic prediction.
• To develop a method that forecasts epidemic trajectories without prior knowledge of all system variables.

Main Methods:

• SPADE4 utilizes a random features model with sparse regression to address data scarcity.
• Takens' delay embedding theorem is employed to reconstruct system dynamics from observable data.

Main Results:

• SPADE4 demonstrates superior performance compared to traditional compartmental models.
• The method's effectiveness is validated on both simulated and real-world epidemic datasets.

Conclusions:

• SPADE4 offers a robust alternative for epidemic forecasting, particularly in data-limited scenarios.
• The approach effectively captures underlying system dynamics for more accurate predictions.