Predicting species interactions (links) within ecological networks is crucial for advancing our understanding of ecosystem functioning and responses of communities to environmental changes. Transductive link prediction models are often used but are constrained by sparse, incomplete data and are limited to single networks. We addressed these issues using an inductive link prediction (ILP) approach to predict interactions within and between ecological networks by pooling data across communities and applying transfer learning. We evaluated the performance of our ILP model on 538 networks across four community types: plant–seed disperser, plant–pollinator, host–parasite and plant–herbivore, and found that it achieved higher precision and F₁ scores than transductive models. However, cross-community prediction efficacy varied, with better performance when plant–seed disperser and host–parasite networks were used as training and test sets, compared with when plant–pollinator and plant–herbivore networks were used. Finally, leveraging the generalizability of ILP, we developed a pretrained model that ecologists could readily use to make instant predictions for their networks. This Article highlights the potential of ILP to improve prediction of ecological interactions, enabling generalization across diverse ecological contexts and bridging critical data gaps.

预测生态系统网络内的物种相互作用（链接）对于促进我们对生态系统功能和群落对环境变化的反应的理解至关重要。经常使用转导链路预测模型，但受稀疏、不完整数据的限制，并且仅限于单个网络。我们使用归纳链接预测 （ILP） 方法解决了这些问题，通过汇集跨社区的数据并应用迁移学习来预测生态网络内部和之间的相互作用。我们评估了我们的 ILP 模型在四种群落类型的 538 个网络上的性能：植物-种子传播者、植物-传粉者、宿主-寄生虫和植物-食草动物，发现它比转导模型实现了更高的精度和 F₁ 分数。然而，跨群落预测效果各不相同，与使用植物-传粉者和植物-食草动物网络相比，当使用植物-种子传播器和宿主-寄生虫网络作为训练和测试集时，性能更好。最后，利用 ILP 的泛化性，我们开发了一个预训练模型，生态学家可以很容易地使用它来对他们的网络进行即时预测。本文重点介绍了 ILP 在改进生态相互作用预测、实现跨不同生态背景的泛化和弥合关键数据差距方面的潜力。