Deployment Optimization of Intelligent Wireless Sensor Network Using Graph Similarity Based on Multi-Granularity Cross Representation and Matching

Intelligent wireless sensor networks often face challenges such as redundancy and non-uniform deployment of sensor nodes, which can negatively impact monitoring performance and energy consumption. To address these challenges, we propose a novel method for identifying the sensor nodes that contribute the most to the monitoring quality of the network. It utilizes graph topology to transform the contribution weights of sensor nodes into the similarity of perturbation-based graphs. And we propose a multi-granularity cross representation and matching method to predict graph similarity, which consists of two stages: representation and matching. In the representation stage, we generate rich multi-granularity interaction features between graph pairs. In the matching stage, we integrate these features into higher-order and more abstract matching features for similarity prediction. To further evaluate the contribution weights of sensor nodes, we combine the obtained graph similarities with the weighted PageRank algorithm. The experimental results demonstrate that our algorithm effectively selects the nodes with greater contribution, leading to good monitoring quality and network performance. Moreover, compared with classical deployment optimization algorithms, the nodes selected by our algorithm are more representative.