Meaning
Algorithmic tests that identify when trajectories in reconstructed state space only appear adjacent due to projecting into an insufficient number of dimensions define false nearest neighbors. By systematically incrementing the embedding dimension and tracking whether neighboring points separate sharply, the procedure determines the minimum coordinate count required to unfold an attractor without geometric overlap. The metric occupies an ordinal scale of integer dimensions bounded at zero, ending its utility once all spurious intersections disappear.
Geometric Unfolding
Dimensional expansion resolves whether proximate historical patterns share identical dynamics or merely intersect through geometric projection. When analyzing business cycles, false nearest neighbors flag scenarios where two historical states look identical simply because crucial operational factors remain unobserved. Adding a coordinate separates points that have no true dynamic affinity, restoring the deterministic path.
The diagnostic eliminates artificial confusion caused by tracking too few operational metrics simultaneously.
Threshold Elimination
Testing neighboring pairs across increasing dimensions uses cardinal distance ratios to mark spurious adjacencies. If the distance between two points expands drastically when moving from dimension d to dimension d plus one, the pair counts as false nearest neighbors. When the percentage of such pairs drops to zero, the embedding dimension suffices to represent the complete dynamic manifold.
At that point, forecasting algorithms operate free of artificial deterministic intersections.
Attractor Resolution
Determining the true integer dimension prevents both under-fitting and unnecessary mathematical overhead in business dynamic models. Low-dimensional models collapse under false nearest neighbors, generating incorrect cyclical forecasts from intersecting historical trajectories. Conversely, excessive dimensions bloat processing costs and introduce observational noise into what should be an orderly deterministic reconstruction.
The exact minimization of false nearest neighbors locks down the natural degree of freedom governing the post-founding firm.
