The world's biodiversity is encountering unprecedented threats, with more than one million plant and animal species at risk of extinction as of 2022. In this thesis, we explore the interactions between plants and arthropods, beginning with an investigation of environmental influences on arthropods using joint species distribution modelling. Our results demonstrate that incorporating specific plants significantly improves model performance, highlighting the importance of considering plant diversity when studying arthropod communities. We then focus on predicting plant-arthropod interactions using Graph Neural Networks (GNN) adapted to our dyadic data. Our proposed approach encompasses different model architectures and choices of arthropod features. We find that GNN-based systems achieve moderate performance, which could potentially be improved by integrating higher-quality data and features. Together, the insights gained from the distributional modelling of species communities and GNN models of dyadic data provide a deeper understanding of plant-arthropod interactions and their complex interplay with diverse factors. This research contributes to the broader knowledge of biodiversity and ecosystem functioning, with potential applications in ecological research and conservation efforts.