{"id":975,"date":"2025-11-18T20:30:32","date_gmt":"2025-11-18T20:30:32","guid":{"rendered":"https:\/\/els-journal.net\/wp\/about\/"},"modified":"2025-11-18T20:30:38","modified_gmt":"2025-11-18T20:30:38","slug":"29204-a-gcn-attention-model-for-precision-irrigation-evaluation","status":"publish","type":"page","link":"https:\/\/els-journal.net\/wp\/?page_id=975","title":{"rendered":"29204 A GCN-Attention Model for Precision Irrigation Evaluation"},"content":{"rendered":"\n\n\n

Vol. 29, No. 2 – December 2025<\/h3>\n

A GCN-Attention Model for Precision Irrigation Evaluation<\/h3>\n

https:\/\/doi.org\/10.53314\/ELS2529070H<\/h5>\n
Ying Huang and Meng Liu<\/h5>\n
Abstract<\/b><\/h5>\n
The challenges of traditional quantitative irrigation methods cannot adapt to the dynamic actual soil moisture content <\/span>and meteorological changes, and the existing methods based on <\/span>soil moisture thresholds cannot fully solve the problems of hysteresis <\/span>and adaptability, lack comprehensive consideration of meteorological <\/span>factors and growth dynamics, and fail to consider the <\/span>subtle sensitivity to soil moisture changes and processing efficiency <\/span>limitations. To address the above challenges, we propose UFOGCN-<\/span><\/h5>
SPANet, a novel and computationally efficient architecture specifically designed for resource-constrained precision agriculture. <\/span>Its core innovation lies in the cascaded integration of: (1) a <\/span>linear-complexity Unit Force Operated Vision Transformer (UFOViT) <\/span>that replaces quadratic self-attention with matrix associativity <\/span>and cross-normalization for efficient global spatio-temporal <\/span>feature extraction; (2) Graph Convolutional Networks (GCNs) for <\/span>modeling spatial dependencies; and (3) a Salient Positions-based <\/span>Attention Network (SPANet) employing a novel Significant Position <\/span>Selection (SPS) algorithm to dynamically focus computation <\/span>on the most informative contextual features, drastically reducing <\/span>complexity while enhancing discriminative power. This unique <\/span>combination directly addresses the critical challenges of computational <\/span>efficiency and effective context modeling in real-world <\/span>irrigation systems. Experimental results show that the proposed <\/span>method outperforms traditional GNN models such as SAGEConv <\/span>with 12 standard time series forecasting methods in key metrics, <\/span>including accuracy, precision, recall, and F1-Score.<\/span><\/h5>\n
Full text: <\/a><\/h5>\n\n\n\n\n\"cited<\/a>\n\n\n\n\nGoogle Scholar Citations <\/a>\n\n\n\n\n
<\/span> <\/center>