arXiv Paper Proposes Auto-ibDLM for Public Event Forecasting via Dynamic Interaction Networks

A recent study published on arXiv (ID: 2608.15488) presents auto-ibDLM, a deep learning framework driven by networks designed for forecasting public events. This framework conceptualizes events as dynamic interaction networks and anticipates their development by estimating participant growth. It employs a hybrid learning approach, initially utilizing network science-informed structural metrics to depict network evolution, followed by converting these structural feature vectors into compact latent representations through an auto-learning layer. A GRU-based module captures temporal dependencies for future participation predictions. The research underscores the significance of accurate public event forecasting for intelligent service systems, facilitating proactive risk management, adaptive resource allocation, and timely decision-making. This approach is inspired by the dynamic interactions among participants in real-world scenarios. The paper is classified as a new announcement and can be accessed via the provided arXiv link.