Determination gender using foot, footprint, hand and hand print measurements in a Sinhalese population in Sri Lanka using supervised learning techniques

Abstract

Foot, footprint, hand and handprint measurements are widely used to determine gender in forensic science to criminal justice and resolve body identification difficulties. Besides, gender determination covers some of the prior identification requirements of a victim. This pioneering study aims to identify the applicability of different machine learning techniques to predict gender using the foot, footprint, hand and handprint measurements in a Sinhalese population in Sri Lanka. Some supervised learning techniques, namely, the Classification and Regression Trees (CART), Naïve Bayes Algorithm and Support Vector Machine (SVM) were used to predict gender. The sample consists of a total of 117 young and healthy undergraduate students, 51 males and 66 females in the age range of 20–30 years. The measurements of foot, footprint, hand and hand print of both left and right sides were measured using standard instruments and techniques. Cross-validation has been used to validate the results. The mean comparison test results show that both left and right side measurements of males are more extensive than that of females (p-value < 0.05). Fitted models show that gender can be determined using the CART algorithm with 95.83% accuracy along with foot length and 91.67% of accuracy along with hand length, hand breadth and palm length. Also SVM, Naïve Bayes and CART algorithms obtained from hand print measurements predict gender with 83.33% accuracy. Particularly, the CART algorithm predicts gender using foot, footprint, hand and hand print measurements with higher accuracies compared to other algorithms. The present study demonstrates the usefulness and reliability of using machine learning techniques for gender determination for the first time among the Sri Lankan population. Hence, the results provide the foundation for future detailed studies.