Archaeologists have to formulate hypotheses and to construct theoretical models in order to explain the social dynamics of past human society and its interactions with the natural world. The large amount of fragmentary data arising from ancient society’s footprints is characterized both by its spatial configuration and by the surrounding environmental conditions. In addition, archeological datasets come from several sources with various variable formats and are therefore often unstructured and affected by noise. The construction of robust and comprehensive analyses for the interpretation of different archaeological contexts is becoming a dominant issue on interdisciplinary research, which
may also include mathematical methods with a spatial statistic imprinting. Quantitative approaches in archaeology are a current trend that significantly highlights the emergence of data-driven based models, with the final goal of extracting as much knowledge as possible form the available information.
The present study deals with ascertained evidences of settlements related to the Roman period inventoried in Switzerland. Archaeological excavations provide a collection of elements characterized by their spatial distribution and additional descriptive information. These can be embedded in a regular grid generating a new collection of regions (e.g. pixels), related to numerical or categorical features (i.e. the presence/absence of settlement and the geo-environmental variables prone to influence their presence). In this study, we constructed a weighted network to model the spatial relationship among the different regions: vertices represent the regions and edges measure the proximity between them, based on the features.