基于元路径的多维信息网络拓扑特征提取仿真

Simulation of Topology Feature Extraction for Multi-dimensional Information Network Based on Meta-Path

针对现有的多维信息网络拓扑特征提取方法存在的特征提取速度较慢、准确性较差等问题,提出基于元路径的多维信息网络拓扑特征提取方法。新方法首先利用有监督特征选择方法获取元路径多维信息网络特征和类标之间的相关性,根据上述相关性完成对多维信息网络特征的选择。然后根据边与类标的相关程度对获取的网络特征进行修剪,此时网络转化为无向无权网络,最后通过计算网络拓扑特征值,完成多维信息网络拓扑特征提取。仿真结果表明,所提方法能够有效的提升拓扑特征提取的速度和准确性。

Due to slow extraction and poor accuracy of existing topology feature extraction methods,this article puts forward a method to extract topology feature of multi-dimensional information network based on meta-path.At first,this method uses the supervised feature selection method to obtain the correlation between the multi-dimensional information network feature based on meta-path and the class label.On this basis,the selection of multi-dimensional information network features was completed.According to the degree of correlation between edge and label,the network features were extracted.At this time,the network was transformed into an un-directed and un-weighted network.Finally,the topological feature extraction of multi-dimensional information network was completed by calculating the network topology feature values.Simulation results show that the proposed method can effectively improve the speed and accuracy of topological feature extraction.