False Nearest Neighbors Method for Calculating Optimal Embedding Dimension

The False Nearest Neighbors method serves as a robust approach for determining optimal embedding dimension in time series analysis. This implementation provides a flexible framework that can be modified according to user specifications. For enhanced customization, users can extend the codebase by adding supplementary modules or refining existing algorithms. Key adjustable parameters include distance metric selection (Euclidean, Manhattan, or custom distance functions) and neighborhood size configuration, which directly impact result accuracy. The algorithm operates by incrementally increasing embedding dimension and identifying false neighbors that disappear with proper dimension selection. Implementation typically involves pairwise distance calculations, threshold-based neighbor identification, and convergence detection mechanisms. This method proves particularly valuable for phase space reconstruction in nonlinear time series analysis, offering researchers a versatile tool for dimensional optimization across various application domains.