基于特征选择的Bagging分类算法研究

Research on Bagging Classification Algorithm Based on Feature Selection

为了提高数据的分类性能,提出了一种基于特征选择的Bagging分类算法。通过Fisher准则和互信息的方法给定一种能够直接评价特征区分度和与类别相关性的评价方法,重新构造了计算特征区分度和与类别相关性的计算公式。并将该方法应用到Bagging分类算法当中。实现了算法迭代过程中的特征选择,使得每个基分类器都是由不同的特征子集训练所得,保证了基分类器的独立性,降低了训练误差。通过理论分析和大量的实验,对文中的方法与经典特征选择方法进行了比较,实验结果显示文中的方法能够得到更高的预测精准度。

In order to improve the classification performance of data,a Bagging classification algorithm based on feature selection is pro- posed in this paper. An evaluation method is proposed for full account of the discrimination and class information of each feature by the Fisher criterion and mutual information ,built on the formula about discrimination and class information. The feature selection algorithm is applied to the Bagging classification algorithm. The feature selection is implemented in the iterative process of algorithm, so that each base classifier is trained by different feature subsets, which ensures the independence of each base classifier, reducing the training error. Com- pared the method with several classical feature selection methods by theoretical analysis and extensive experiments, the results show that the method can achieve higher predictive accuracy.