For a study, researchers sought to clarify the dynamic features that are highly predictive in the biological and perceptual sex classification of point-light walkers (PLWs) as how these features behave in sex classification using supervised machine learning. 

From a side perspective, 15 observers assessed the sex of 21 PLWs. A quick Fourier transform was used to extract the spectral components from the multiphasic hip and shoulder motions. The most crucial characteristics for biological and perceptual sex classifications were found after a thorough examination. To explain the behavior of each significant feature, a support vector machine (SVM) model and an individual conditional expectation (ICE) were utilized. 

The observers correctly identified the biological sex in 10 male and 11 female side-view PLWs with an accuracy of 62.9% and 57.0%, respectively. The third harmonic of hip motion had a prominent role in generating a high predictive accuracy of 90.5% with few feature interactions, according to the SVM model for biological sex prediction. While the ICE plots of the features in the model of perceptual sex prediction followed diverse paths, indicating feature interactions, an accurate prediction of 85.7% was made utilizing five spectral components of hip and shoulder movements. 

According to the machine learning model, PLW local cues play a major role in biological sex categorization. Though interactions between different frequency components of hip and shoulder movements were included in the high-performance model of perceptual sex categorization, it suggested more intricate mechanisms in sex perception.