具有更好适应性的间距最大化特征加权

Margin maximization feature weighting with better adaptability

间距最大化特征选择技术是一种有效的维数约减技术,一般是基于加权技术和相似性度量构造目标函数。不足之处是针对不同数据集,适应性有待提高。为此,引入一个具有更好适应性的度量(称为分离度)和模糊加权技术构造新目标函数。利用模拟数据集和基准数据集进行仿真,实验结果表明该方法具有良好的适应性。

Margin maximization feature weighting is an effective dimension reduction technique, and it is generally based on weighting techniques and similarity measure to construct their objective functions. One of the weaknesses is that the adaptability is not enough due to the complexity of different datasets. Therefore, the better adaptive margin metric （ named separation） and the fuzzy weighting technique were introduced for developing the new optimal objective function. As demonstrated by the authors＇ experimental studies in synthetic datasets and benchmark datasets, the proposed algorithm is more adaptive.