K-Means Clustering for Evolutionary Staging in a Human Evolution Dataset

Abstract

This research work applies unsupervised machine learning to explore evolutionary patterns in hominin morphological and temporal data. A dataset comprising 6,000 records of hominin specimens was analysed using three quantitative attributes: geological age (1–8 million years), cranial capacity, and estimated stature. Following data cleaning and z-score normalization, K-means clustering (K = 4) was employed to identify coherent evolutionary groupings without prior taxonomic labelling. The resulting clusters exhibit a clear temporal and morphological progression. The earliest cluster (mean age ≈ 6.66 Ma) is characterized by the smallest cranial capacity (≈156 cm³) and stature (≈106 cm), consistent with early hominin forms. A second cluster (≈3.89 Ma, 367 cm³, 117 cm) corresponds to Australopithecine-like specimens, while a transitional cluster (≈1.96 Ma, 490 cm³, 119 cm) reflects early Homo characteristics. The most recent cluster (≈1.07 Ma) displays substantially larger cranial capacities and statures (≈1063 cm³ and ≈162 cm), aligning with later or near-modern Homo. Visualization through scatter plots, bar charts, and boxplots supports a monotonic increase in cranial capacity and height across evolutionary stages. These findings demonstrate that unsupervised clustering can recover biologically meaningful evolutionary patterns from morphological and temporal data, highlighting its potential as an exploratory tool in paleoanthropological research.