基于模糊互信息的多标签特征选择

Multi-label Feature Selection Based on Fuzzy Mutual Information

多标签特征选择已成为处理多标签数据的重要方法之一.利用一种基于模糊互信息的多标签特征选择算法,通过模糊离散化,给出模糊联合熵和模糊条件熵吸模糊互信息的计算方式.将原先的互信息特征选择算法推广到模糊情形中,提出一种基于模糊互信息的多标签特征选择算法.最后在同一准则下,将模糊互信息与原先的互信息特征选择算法及其他几种经典的特征选择算法进行比较.实验表明,此方法在一定程度上效果优于其他特征选择算法,是一种有效的多标签分类问题的特征选择方法.

Multi-label feature selection is one of the most important methods for multi-label data processing.A multi-label feature selection algorithm based on fuzzy mutual information is proposed.Firstly,the fuzzy mutual information between attribute sets is considered by the label set,and the fuzzy joint entropy and fuzzy conditional entropy between attributes are taken into account.Then,the original mutual information feature selection algorithm is extended to the fuzzy case,and a new method based on fuzzy mutual information is proposed.Finally,under the same criterion,the fuzzy mutual information is compared with the original mutual information feature selection algorithm and several other classical feature selection algorithms.Experiments show that this method outperforms other feature selection algorithms to a certain extent and it's an effective feature selection method for the multi-label classification problems.