In forensic and archaeological contexts, accurately determining an individual's age, sex, height, and weight is essential for the identification of unidentified bodies and dismembered remains. This study assesses the effectiveness of linear regression analysis and artificial neural networks (ANNs) in estimating body weight using foot measurements within the Eastern Turkish population. The research was conducted with medical students, including 149 volunteers—76 males and 73 females. Participants' height and weight were recorded using a stadiometer. Foot measurements were taken with a Vernier caliper and an osteometric board. The data were assessed utilizing SPSS 26.0, and ANN models were developed using the Keras library within the Google Colab platform. In terms of height, weight, and foot size, all these values were significantly higher in males than in females. Linear regression methods resulted in a root mean squared error (RMSE) of 11.83 kg for males, 6.23 kg for females, and 9.56 kg for the entire sample. With ANNs, these error values were reduced to 10.13 kg for males, 5.10 kg for females, and 8.07 kg for the entire sample. The analysis shows that ANNs are more adept at handling complex data, leading to improved prediction accuracy.
Key words: Weight estimation, anthropometry, forensic anthropology, linear regression, artificial neural networks
|
