An Evaluation of Link Prediction Approaches in Few-Shot Scenarios

• DOI: 10.3390/electronics12102296
• Publication Date: 2023-05-19T13:23:10Z
• Authors: Rebecca Braken; Alexander Paulus; André Pomp; Tobias Meisen
• Keywords: Link (geometry)

Semantic models are utilized to add context information datasets and make data accessible understandable in applications such as dataspaces. Since the creation of is a time-consuming task that has be performed by human expert, different approaches automate or support this process exist. A recurring problem link prediction, i.e., automatic prediction links between nodes graph, case semantic models, usually based on machine learning techniques. While, general, trained evaluated large reference datasets, these conditions often do not match domain-specific real-world wherein only small amount existing available (the cold-start problem). In study, we performance algorithms when smaller size were used for training (few-shot scenarios). Based reported evaluation, first selected then subset using multiple reduced datasets. The results showed two three suitable few-shot scenarios.