Lecture | Digital Archaeology Group
Analysis of clustering algorithms and performance evaluation metrics applied to samples of the Tell El-Yahudiya ware typology
- Date
- Thursday 24 April 2025
- Time
- Address
- Hybride
- Room
- Van Steenis, C1.24
This paper is related to a PhD research on pattern recognition methods, especially unsupervised learning, combined with analytical tools and interpretive approaches for constructing pottery typologies in archaeology. In the first stage of this research, different clustering algorithms are applied to a pottery dataset with an already defined chrono-cultural typology, to evaluate the potential that algorithms built with different concepts and several combinations of parameters have to replicate the original expert typology. Several combinations of features are taken into account. Clustering performance is being evaluated through both ground-truth based metrics and model-only based metrics.
The lecture can be attended on campus or online via https://digarchgroup.nl/lecture.