R is a free statistical programming language. It is extensively used in many disciplines, including Archaeology: it became a mighty tool to approach the growing amount of quantitative data in archaeological research. On 6th February 2019, at the third day of the 'International Colloquium on Digital Archaeology in Bern', the participants had the opportunity to attend a R workshop where they worked on real archaeological data and learned the application of machine learning techniques.
In particular, we focussed on predictive modelling, or more precisely predictive mapping, as a prediction tool for archaeological site potential. The methodological basis of this application was briefly discussed and statistical tools from the field were introduced shortly. Then, we intersected an archaeological data set (site information) with spatial environmental data and estimated and mapped the site potential using two methods (Generalized Linear Modelling and Naïve Bayes Classifier). The plan was to compare both results and to discuss the problem of overfitting machine learning procedures.
Lecturer: Dr. Martin Hinz, University of Bern
Knowledge in R programming is of advantage, but not absolutely necessary. No actual introduction in R is given. At the end of the tutorial, participants should be able to perform predictive mapping.
Especially towards the end (Part III), the presentations are still not perfectly layouted. This is due to the fact that ad-hoc changes took place. I will improve this part successively.
Here I will collect a list of literature that might be helpful to understand the background of archaeological predictive mapping. To be filled...