摘要
生态系统功能依赖多层级交互维持,生态网络以图结构表达种间关系与耦合机制,其动态行为体现非线性、异质性与时变性特征。文章构建基于图神经网络的复杂生态网络动力学建模方法,涵盖异构节点编码、动态边权生成与环境因子嵌入,提出结构与时间解耦的注意力机制与多任务优化策略。文章结合红海—沙漠生态交错区样区监测数据,构建时序图结构,量化评估模型在状态预测、边权拟合与预警响应等任务中的表现,验证所提方法在生态智能感知中的可行性与扩展能力。
Ecosystems rely on multi-level interactions to maintain their functions.While ecological networks express inter species relationships and coupling mechanisms through graph structures.Their dynamic actions are nonlinear,heterogeneous and time-varying.To this end,the article proposes a dynamic modeling method for complex ecological networks based on graph neural networks,covering heterogeneous node encoding,dynamic edge weight generation,and environmental factor embedding.Then,it proposes an attention mechanism and multi-task optimization strategy that decouples structure and time.In the end,drawing on monitoring data from the Red Sea-Desert ecological ecotone sample area,a time-series diagram was constructed to evaluate the performance of the model in tasks such as state prediction,edge weight fitting,and warning response,to verify the feasibility and scalability of the proposed model in ecological intelligent perception.
作者
侯天芳
Hou Tianfang(Sanmenxia Polytechnic,Sanmenxia 472000)
出处
《中阿科技论坛(中英文)》
2025年第8期79-83,共5页
China-Arab States Science and Technology Forum
关键词
图神经网络
生态网络建模
动力学演化
时空注意力机制
Graph neural network
Ecological network modeling
Dynamic evolution
Spatiotemporal attention mechanism
