Machine learning-based approaches for predicting stature from archaeological skeletal remains using long bone lengths

• We are approaching the problem of estimating the stature of human skeletal remains.
• Two novel machine learning based approaches are proposed for the considered problem.
• The experimental evaluation confirms that our approaches outperform the existing ones.

This paper approaches, from a computational perspective, the problem of predicting the stature of human skeletal remains from bone measurements. There are traditional methods for constructing models that give good results for stature estimation. In this paper, we aim to investigate the usefulness of using machine learning-based models to approximate stature. Assuming that the stature of an individual is indirectly related to bone measurement values, we can derive methods that learn from archaeological data and construct models that give good estimates of the stature. Two novel machine learning-based regression models for stature estimation are proposed in this paper. Experiments using artificial neural networks and genetic algorithms were performed on samples from the Terry Collection Postcranial Osteometric Database, and the obtained results are discussed and compared with the results from other similar studies. The experimental evaluations indicate that the machine learning-based regression models are efficient for the stature estimation of archaeological remains and highlight the potential of our proposal.