Cluster Analysis of Features Associated with Unidentified Anomalous Phenomena Described in 216 Select Reports from 1947–2016

Witness reports of unidentified aerial phenomena (UAP) contain descriptions of observed UAP characteristics and behaviors. We sought to gain insights into possible classes of UAP by applying two-step cluster analysis, a statistical pattern recognition technique, to data extracted from the select UAP report database (n = 301) created by Powell et al. To create the clusters, we targeted shape, estimated size, ability to hover, presence of electromagnetic (EM) effects, and presence of sound, using the subsample (n = 216) with complete data for these characteristics. We identified a model of good statistical quality (silhouette value = 0.6) with seven distinct clusters. Object shape was the primary driver of clustering results overall (disk, cylinder, triangle), however some distinct clusters were dominated by: absence of hovering (one disk; one mixed shape); presence of sound (mixed shape); and EM effects (disk/sphere or ovoid). Notably, of the objects exhibiting EM effects, 97% were viewed at estimated distances of ≤ 2,000 feet from the observer. Findings suggest that statistical pattern recognition techniques like cluster analysis augment results of manual grouping of UAP types. Such techniques can extract subtle and meaningful information from high quality witness report data that will ultimately lead to an improved understanding of the nature of UAP.