Gender Estimation from 2D:4D Ratio and Hand Morphometry by Using Machine Learning Algorithms

Abstract

Background: The present study was conducted to estimate gender from 2D:4D ratio and hand mor-phometry taken from participants by using machine learning (ML) algorithms. Materials and Methods: The study was conducted retrospectively on 88 men and 96 women be-tween the ages of 18 and 30 who did not have any pathology, deformity or surgical interventions on their hands. Hand width (HW), hand length (HL), second digit length (2D), and fourth digit length (4D) of the individuals were measured as the right (R) and left (L) side by using digital calliper and recorded in Excel. In addition, the ratio between the second digit and fourth digit (2D:4D) of each individual was also recorded. Results: As a result of ML modelling, 0.92 accuracy was obtained with Random forest (RF) algorithm. With RF algorithm, all of the 16 women and 18 of the 21 men in the test set were estimated accu-rately. With SHAP analyzer of RF algorithm, HW-L parameter was found to have the highest contri-bution in estimating gender. The accuracy rates of the other ML models used in the study were found to vary between 0.78 and 0.89. Conclusions: It was found that 2D:4D ratio and hand morphometry measurements, which are fre-quently preferred in gender determination, have higher accuracy rate when examined with ML algorithms. In our study, we concluded that using 2D:4D ratio and hand morphometry in estimating gender provides accurate and reliable data.