基于超长方体与KNN的分类算法

Classification Algorithm Based on Multiple Enclosing Hyper Cubes and KNN Method

提出了超长方体与KNN相结合的分类算法。在训练阶段,该算法为训练集中的每一个类别构造多个超长方体,区域分离每一类训练样本。在测试阶段,该算法首先检查测试样本是否被某一个超长方体包围,如是则其类别被识别出,否则用KNN方法确定其类别。实验采用四个真实数据集进行测试。实验结果表明基于超长方体与KNN的分类算法在四个数据集全部优于两个基于多球覆盖的分类方法,是一种有效的分类方法。

This paper proposes a classification algorithm based on the multiple enclosing hyper cubes and the KNN method. This algorithm builds many hyper cubes to enclose each type oftraining samples in the separated regions. In the stage of cube. The type will be confirmed tion. Four real datasets are used t performs better than two kinds of method in three data sets. This al test, every testing sample is checked whether it is enclosed by a hyper if it is this case. Otherwise, KNN method is used to realize classifica- o do the experiments in this paper. The results show that this algorithm hyper sphere-based classification methods and surpasses their ensemble gorithm is an effective classification method.