一种面向标签排序数据集的特征选择方法

针对标签排序问题的特点，提出一种面向标签排序数据集的特征选择算法（Label Ranking Based Feature Selection, LRFS）。该算法首先基于邻域粗糙集定义了新的邻域信息测度，能直接度量连续型、离散型以及排序型特征间的相关性、冗余性和关联性。然后，在此基础上提出基于邻域关联权重因子的标签排序特征选择算法。实验结果表明，LRFS算法能够在不降低排序准确率的前提下，有效剔除标签排序数据集中的无关特征或冗余特征。

A feature selection method for label ranking dataset

This paper proposed a feature selection method according to the characteristic of label ranking problem. First, some new neighborhood information measures are defined based on the neighborhood rough set which can measure the relevance, redundancy and interaction of different kinds of features directly. Then, a feature selection algorithm is put forward based on the neighborhood interaction weight factor. Experimental results show that the algorithm can effectively eliminate the irrelevant features and redundant features of label ranking dataset without reducing the ranking accuracies.