Osteoarcheology Research Papers - Academia.edu (original) (raw)

In recent years, the application of new data analysis techniques due to the rise of the computational power available for a much larger number of people has caused a revolution inseveral fields, giving the opportunity of searching new approaches to familiar issues.Machine learning (ML) is one of those approaches that is rising in popularity, even when the theoretical base and first applications have been developed a number of decades ago.Generally speaking, ML describes a kind of algorithms that are able to learn on their own, being able to make models and predictions after having trained with a sample of the data(Domingos, 2012). This can be especially useful in classification problems (Smola Vishwanathan, 2008).ML has been applied in archaeology before, for example in the creation of complex models about the human past (R. Vahdati et al., 2019), But it is its interesting capacity for imageprocessing that has been more used in this field, using drone imagery for automated surface surveying (Orengo GarciaMolsosa, 2019), satellite imagery for archaeological research inareas with difficult access (Davis Douglass, 2020), an automatic system of medieval coin classification (van der Maaten et al., 2006), or aerial imagery for mapping of archaeologicalstructures (Monna et al., 2020).Sex-determination of human remains, on the other hand, is one of the most common classification problems that can be found in archaeology, and specifically, in osteoarchaeology, asin other sciences too, such as forensic science or physical anthropology (Krishan et al., 2016). Standard anthropometric measurements have been used in those disciplines from theXIXth century and the relationship between those measurements and the size, sex and general build of the person where discovered (Krishan, 2006).The purpose of this study is to explore how suitable the application of different ML algorithms is to tackle the classical classification problem of sex determination in human skeletalremains using long bone anthropometric measurements can be and if it could be used in some of the possible bone conservation scenarios that can be found in field archaeology,and how high can the accuracy reached for these techniques be when applied to a general population. The aim is to develop general classification models that could be applied inarchaeological, anthropological or forensic scenarios when the specific origin of the remains is unknown