基于子图交互关系的网络结构增强算法

已有研究基于子图交互关系构造子图网络来实现网络结构增强，然而其算法复杂度高。鉴于此，基于不同阶子图网络的拓扑属性分别对原始网络进行赋权，得到一阶和二阶加权网络，以权重的形式直观体现子图交互关系。同时，这两种加权网络的权重可以直接通过原始网络的拓扑结构计算得出，从而避免了子图网络的构造过程，大大降低了算法复杂度。最后，以关键点识别任务作为研究对象说明这两种加权网络在结构挖掘应用中的性能。基于加权网络定义了两种新的中心性指标，在8个真实网络中与7种经典的中心性指标进行对比，实验结果表明基于加权网络的中心性指标具有更好的性能。

Network Structure Enhancement Algorithm Based on Subgraph Interaction

The existing studies show that the network structure can be enhanced by constructing subgraph network based on subgraph interaction relationship, but the complexity of such algorithms is high. In view of this, this paper weights the original network based on the topological attributes of different order subgraph networks, obtains the first-order and second-order weighted networks, and intuitively reflects the interaction relationships of subgraphs in the form of weights. At the same time, the weights of the two weighted networks can be calculated directly through the topology of the original network, which avoids the construction process of subgraph network and greatly reduces the complexity of the algorithm. Finally, the key nodes identification task is taken as the research object to illustrate the performance of the two weighted networks in the application of structure mining. In this paper, two new centrality indices are defined based on weighted networks, which are compared with seven classical centrality indices in eight real networks. The experimental results show that the centrality indices based on weighted network has better performance.