Luca Deravignone / Giancarlo Macchi

((ASIAA lab (Laboratory of Spatial Analysis and Information Technology Applied to Archaeology), University of Siena, Italy)

University of Siena. The aim of this paper is to illustrate some of the results achieved in the study of medieval settlement patterns in southern Tuscany between the XII and the XIV century, through the use of quantitative techniques and spatial analyses based on the application of Artificial Neural Networks (ANN). In this effort ANN had been used to estimate and analyze correlation between fortified villages locations and settlements systems related to other historical phases. Significance of quantitative grids had been improved by evaluating and matching other environmental and ecological variables like morphology or distances from natural resources.
All this, trying to answer how ANN changed after a decade of studies traditional paradigms concerning relationships between human settlements with previous and successive systems in terms of background and feedback. Each village can be seen as an outcome of a background, but also as a feedback for future settlement development. In the same way, ANN had been used to measure differences in the medieval settlement system inside the Tuscany boundaries between different districts. Besides, another objective of the project is to achieve a geographically-based classification of the settlements categories from late antiquity to the end of the middle ages between southern Tuscany and northern Lazio.
Such tasks required the development of a large number of dedicated software. These tools where conceived and developed for an intuitive and automated application in the analysis process. Among these an ArcGIS plug-in that allow the final user to generate all the required files to train ANN inside the SNNS (Stuttgart Neural Network Simulator), but also to handle and apply trained ANN within GIS grid analysis routines.