Looking for the Best. A Comparison Between GIS and PageRank Based Algorithms for Preventive Archaeology in Urban Areas

In this paper we present a comparison between the MAPPA algorithm and some more standard predictive models implementable within the GIS: kernel density estimation (KDE), kriging/co-kriging and r.finder. The MAPPA algorithm is a mathematical model, based on PageRank, that imitates the modus operandi of archaeological practice, reproducing the relations between the different natural and anthropogenic elements. Kde is a non-parametric way to estimate the density of a random variable, in which a function creates a continuous approximated distribution. Kriging is an interpolation method, based on spatial autocorrelation. Co-kriging allows to add further spatial information in order to improve the interpolation. r.finder is a GRASS-GIS script created to check the analogies among the places where some known items are located, in terms of similar cell-values in the frame of a series of thematic raster maps. The obtained results confirm the better suitability of predictive models specifically developed by and for archaeologists.