A PageRank based predictive model for the estimation of the archaeological potential of an urban area

We present the analysis of multi-faceted, GIS managed data for determining the archaeological potential, i.e. a measure of the possibility that a more or less significant archaeological stratification is preserved. We used a sizable number of datasets, in order to consider the problem of estimation of archaeological potential in all of its aspects: archaeological data, building archaeological data, historical data, toponymic data, geomorphological data. As the identification of relations among finds is a key issue for the data mining in archaeological interpretation process, we applied a modified version of the PageRank model, because the criteria for assigning importance to web pages by search engines are similar and based on relations, also. The procedure included a categorization archaeological data, the assignment of initial values of potential to the available data through an automatic procedure, the creation of geomorphological facies maps, the definition of functional areas (i.e. the levels of spatial and functional organization: urban, suburban and rural areas), and the application of the PageRank based algorithm. The model has been applied on the urban area of Pisa, and tested through the data of 14 new cores. The map of archaeological potential consists of the composition of the 7 layers, one for each archaeological period under consideration: Protohistory, Etruscan period, Roman period, Late Roman period, Early Medieval period, Late Medieval period, Modern Age, Contemporary Age. The results, including the archaeological potential map, are to be considered as the first steps towards an automatic, formally definable, and repeatable, approach to the computation of archaeological potential.

I. Introduction

This paper studies the problem of computation of archaeological potential, the assumptions made to solve it, the mathematical model used, the software implementation, and the test of the algorithm in the case study of the urban area of Pisa. We based the mathematical model on PageRank, because there is an analogy between the criteria used for attributing archaeological potential and the criteria used for assigning importance to web pages in search engine algorithms. The key issue of the computation of archaeological potential, from an abstract viewpoint, is the identification of relations among finds: the presence of a particular find near another could strengthen or weaken the probability that they will form a more complex structure, and so strengthen or weaken the archaeological potential of the area. This is exactly the criterion upon which page ranking algorithms are based, whereby each web page attributes importance to the web pages it points to (via a link) and receives importance from the web pages it receives a link from. The reader can refer to [4] for further explanations about the choice of the mathematical model, and to [8] for a general mathematical introduction to PageRank models. In the following we will consider all the archaeological data as categorised, having assigned each find to a category in order to characterize its salient features, to effectively implement the algorithm, and to make the results general enough to be applied also in different contexts (pp. 89–99, [2]).