Optimal Selection of Parameters for Nonuniform Embedding of Chaotic Time Series Using Ant Colony Optimization

Area of Science:

• Complex Systems Analysis
• Time Series Forecasting
• Computational Intelligence

Background:

• Optimal parameter selection is critical for analyzing and forecasting chaotic time series.
• Existing methods for uniform embedding parameter selection are well-developed, but nonuniform embedding lacks efficient parameter selection techniques.
• Nonuniform embedding, allowing varied time delays across dimensions, poses a combinatorial optimization challenge.

Purpose of the Study:

• To propose an efficient Ant Colony Optimization approach for selecting parameters in nonuniform embedding.
• To optimize both the embedding dimension and time delays simultaneously for chaotic time series analysis.
• To enhance the accuracy of time series forecasting using optimized nonuniform embeddings.

Main Methods:

• Developed an Ant Colony Optimization for Nonuniform Embedding (ACO-NE) algorithm.
• Divided the solution construction into two phases: embedding dimension selection and time delay selection.
• Incorporated heuristics derived from time series data to guide the ant colony search and accelerate convergence.

Main Results:

• ACO-NE successfully optimizes embedding dimensions and time delays for nonuniform embedding.
• The algorithm demonstrates efficient search capabilities by utilizing heuristic information.
• Embeddings generated by ACO-NE were effectively applied to time series forecasting tasks.

Conclusions:

• ACO-NE provides an efficient and effective method for solving the combinatorial optimization problem in nonuniform embedding.
• The proposed approach yields superior embedding solutions compared to existing methods.
• Optimized embeddings derived from ACO-NE lead to improved prediction accuracy in chaotic time series forecasting.