Quantitative methods have long been employed in archaeology, yet the application of machine learning techniques to archaeological data has only recently gained momentum. This paper presents an example of using machine learning to enrich datasets from the Late Antique cemetery Lajh in Kranj. Of 544 graves analysed, only 88 (16.2%) have anthropological sex determinations, largely due to early excavation practices where skeletal remains were not systematically preserved. To address this gap, we trained a logistic regression model on 69 graves characterised with both grave goods and anthropologically
determined sex. Using the presence and quantity of 68 categories of grave goods, the model predicted sex labels for all the graves with grave goods and provided probability scores for each prediction.
Reliable predictions (> 75% probability) were obtained for 188 graves, effectively expanding the dataset for further analysis.The model revealed meaningful patterns, for example, those associating specific grave goods with gender and the expression of gender in children’s burials. While the approach is limited by data quality, typological generalisations, and small training sets, it demonstrates how machine learning can highlight relationships between variables and provide additional perspectives on gender in mortuary contexts.
|
