基于模糊测量函数的多标签图数据特征提取研究

Research on feature selection of multi-label graph date based on fuzzy measure function

如今,图数据分类变得越来越重要。最近几十年对它的研究也越来越多,并且得到了广泛应用。传统的图数据分类研究主要集中在单标签集,然而在很多应用中,每个图数据都会同时具有多个标签集。这篇文章研究了关于图数据的多标签特征提取问题,并提出基于模糊测量函数的多标签图数据特征提取算法,用于搜索最优子图集。算法采用模糊测量函数作为评估标准评估子图特征的重要性,然后通过边枝界定算法修剪子图搜索空间有效地搜索最优子图特征。实验证明,该方法在现实应用中有较高的精度。

Nowadays,the graph data classification has become more and more important. In recent decades,the research on it is also more and more,and has been widely used. The traditional graph data classification research mainly focused on single label set,however,in many applications,there will be multiple label set at the same time in every graph data. This paper studied the problem of multi-label feature extraction of graph classification and proposed a feature selection algorithm of multi-label graph date based on fuzzy measure function,which is used to search the optimal sub graph set. The algorithm adopted fuzzy measurement function as evaluation standard to evaluate the importance of sub graph feature,and through the side branches defining algorithm pruned graph search space to effectively search the optimal sub graph features. Experiments showed that the method has high accuracy in practical applications.