Classification archaeological fragments using data mining techniques

The archaeological record produces a vast yet discontinuous collection of information, reflecting fragmented traces of human activity across time and space. Understanding these discontinuities is vital not only for accurate dating but also for interp reting cultural evolution and facilitating comparative archaeological studies. In this context, the paper proposed the use of self-supervised learning as a method for feature fusion and classification techniques, emphasizing multimodal and bias-aware approaches. This framework allows the integration of diverse data types such as imagery, stratigraphic data, and contextual metadata into a coherent model capable of recognizing complex patterns. The findings of the probabilistic derived from this approach demonstrate the robustness and interpretative power of computational archaeology, revealing how machine learning can enhance and develop traditional archaeological reasoning. Ultimately, such techniques provide general archaeologists with stronger analytical tools for characterizing datasets, addressing uncertainty, and making reliable inferences about cultural processes even in out-of-sample or incomplete data scenarios.