Ecological network models are essential for developing and quantifying ecosystem-based management strategies. Unobserved species interactions alter the interpretation of structural and functional characteristics of the ecosystem being studied. Link prediction algorithms can help to identify such unobserved, ‘hidden’ interactions. However, due to general unfamiliarity and insufficient ecological interpretations, the use of link prediction algorithms in ecology remains limited. In this study, we enhance the link prediction applicability in ecological networks by considering and quantifying the algorithm results from the link as well as the node perspective using a coastal food web model from the northern Wadden Sea as a case study. For this purpose, we have defined the Weighted Unobserved Node Connectivity (WUNC) representing a new node property. The WUNC facilitates the estimation of the missing connectivity of a species in relation to a considered original source network. Such a new combination of both link and node perspectives helps to uncover unobserved interactions as well as the resulting lack of species connectivity in poorly understood environments without active sampling. The bi-dimensional perspective presented in this study provides a more effective use of link prediction algorithms to identify and prioritize under-connected species and their unobserved interactions. This enables the design of more targeted, species-specific measurement campaigns to validate predicted interactions, thereby supporting refinements of existing ecological network models. A more comprehensive representation of interactions in ecological network models contributes to more accurate modelling results and improves their interpretation to support better management strategies in times of environmental changes.