Analytical archaeology: novel techniques in archaeological predictive modeling using AI and machine learning

Keywords: Predictive modeling, Archaeology, Machine Learning, Artificial Intellegence
Session Type: Virtual Guided Poster Abstract
Day: Monday
Session Start / End Time: 2/28/2022 11:20 AM (Eastern Time (US & Canada)) - 2/28/2022 12:40 PM (Eastern Time (US & Canada))
Room: Virtual 49

Abstract

The field of predictive site modeling in archaeology employs various tools to predict the probability of archaeological site presence at particular locations on the landscape given a set of known site locations, non-site locations, and environmental and geophysical variables such as distance to water, elevation, or native vegetation. In this research, we revisit several archaeological predictive modeling case studies. We apply analytical techniques rooted in GIS and machine learning for each case study, specifically TensorFlow neural networks. We identify the extent to which these strategies offer improvements for archaeological site prediction or not. Analytical archaeology rooted in AI and machine learning appears to provide substantial gains over traditional methods, including improved predictive accuracy and substantially faster model fitting times when compared to more traditional methods such as Bayesian logistic regression. With advances in cheaper, more powerful computing, improved environmental datasets, and ever-growing communities of data analysts, we evaluate to what extent we can improve our understanding of the past using these cutting-edge technologies. The research here identifies potential technological and analytical gains when these methods are applied to archaeological problems, likely outcomes, difficulties encountered along the way, and future directions in this area of research.