中文摘要：

k子凸包分类方法在实际问题中有广泛应用.但是该方法仍然对噪声和参数k比较敏感,并且在k邻域内不同类的样本数经常严重失衡,导致分类性能下降.针对上述问题,设计了一种选择性自适应k子凸包分类方法.首先根据k子凸包分类的特点给出冗余数据、噪声和决策邻域的概念,并对数据进行网格化处理.然后采用留一法对数据集进行选择性修剪,去掉冗余数据和噪声;并为每个样本学习一个不同的决策邻域,使得不同样本的决策邻域能够自适应变化.实验表明,该方法不仅缩小了问题规模,而且分类性能也有显著提高.

英文摘要：

The k sub-convex-hull classifier is widely used in the practical problems.But this method is still quite sensitive to the noise and the parameter k.Moreover,different types of samples in k-nearest neighbors of a test instance often result in serious imbalance,leading to the decline of classification performance.In this paper,we propose a selective adaptive k sub-convex-hull classifier（SACH）to address these problems.Firstly,we give the definition of redundant data,noise,and decision-making neighborhood according to the characteristics of the k sub-convex hull classifier.Meanwhile,in order to effectively deal with large data sets,a large data set is divided into many small data grids through the application of grid technology.Then,we use the leave-one-out to prune selectively data sets,and remove redundant data and noise.Moreover,the decision neighborhood of each data can adaptively change by learning different decision neighborhood boundaries for each data.Experimental results show that our selective adaptive k sub-convex-hull classifier not only can reduce the size of the problem,but also can significantly improve the classification performance of the k sub-convex-hull classifier.