Predicting the number of new relationships based on combinations of people is a helpful task for decisionmaking.This paper proposes an Interpretable Relation Prediction (IRP) framework to predict the complex relationships between individuals and groups. Our two-stage framework uses a dynamic bipartite graph as an input to simplify increasing sets of nodes over time. The framework's high interpretability in handling temporal dynamics and intuitive features provides insights into the underlying factors driving the predictions, making it useful for decision-making in various domains, such as recommendation systems, social networks, and academic co-authorship networks.