基于相关性分析及遗传算法的高维数据特征选择

特征选择是模式识别及数据挖掘等领域的重要问题之一。针对高维数据对象，特征选择一方面可以提高分类精度和效率，另一方面可以找出富含信息的特征子集。针对此问题，提出了一种综合了filter模型及wrapper模型的特征选择方法，首先基于特征与类别标签的相关性分析进行特征筛选，只保留与类别标签具有较强相关性的特征，然后针对经过筛选而精简的特征子集采用遗传算法进行随机搜索，并采用感知器模型的分类错误率作为评价指标。实验结果表明,该算法可有效地找出具有较好的线性可分离性的特征子集，从而实现降维并提高分类精度。

High-dimensional data feature selection based on relevance analysis and GA

Feature selection is one of the important problems in the pattern recognition and data mining areas. For highdimensional data feature selection not only can improve the accuracy and efficiency of classification, but also can discover informative feature subset. The new feature selection method combining filter and wrapper models was proposed, which first filters featured by feature relevance analysis, and realized the near optimal feature subset search on the compact feature subset by genetic algorithm; and the feature subset was evaluated by the classification inaccuracy of the pereeptron model. The experiments show that the proposed algorithm can find the feature subsets with good linear separability, which results in the low-dimensional data and the good classification accuracy.