Sujet Design, implementation and analysis of a description model for complex archaeological objects
Date lundi 9 juillet 2018 à 14h30
Lieu Université Lyon 2, Salle A103, Campus Porte des Alpes, 160 Boulevard de l'université, 69500 Bron
Ceramics are one of the most important archaeological materials to help in the reconstruction of past civilizations. Information about complex ceramic objects is composed of textual, numerical and multimedia data, which induce several research challenges addressed in this thesis. From a technical perspective, ceramic databases have different formats, access protocols and query languages. From a data perspective, ceramic data are heterogeneous and experts have different ways of representing and storing data. There is no standardized content and terminology, especially in terms of description of ceramics. Moreover, data navigation and observation are difficult. Data integration is also difficult due to the presence of various dimensions from distant databases, which describe the same categories of objects in different ways.
Therefore, the research project presented in this thesis aims to provide archaeologists and archaeological scientists with tools for enriching their knowledge by combining different information on ceramics. We divide our work into two complementary parts: (1) Modeling of Complex Archaeological Data and (2) Clustering Analysis of Complex Archaeological Data. The first part of this thesis is dedicated to the design of a complex archaeological database model for the storage of ceramic data. This database is also used to source a data warehouse for doing on-line analytical processing (OLAP). The second part of the thesis is dedicated to an in-depth clustering (categorization) analysis of ceramic objects. To do this, we propose a fuzzy approach, where ceramic objects may belong to more than one cluster (category). Such a fuzzy approach is well suited for collaborating with experts, by opening new discussions based on clustering results.
We contribute to fuzzy clustering in three sub-tasks: (i) a novel fuzzy clustering initialization method that keeps the fuzzy approach linear; (ii) an innovative quality index that allows finding the optimal number of clusters; and (iii) the Multiple Clustering Analysis approach that builds smart links between visual, textual and numerical data, which assists in combining all types of ceramic information. Moreover, the methods we propose could also be adapted to other application domains such as economy or medicine.
Complex Objects, Archaeology, Archaeometry, Ceramics, Databases, Data Warehouses, OLAP, Clustering, MaxMin Linear Initialization, Cluster Validity, Visual TSFD, Cluster Ensemble, Combined Partition Clustering