适用于关联属性的样本自适应参数孤立点检测法

为解决数据集中关联属性之间的干扰问题,通过引进Mahalanobis距离,并对传统的k近邻孤立点检测方法进行了改进,提出了一种新的基于样本的参数选取方法。该方法通过训练数据集中的正常数据和孤立点数据,以获得最优的k距离值和阈值。实验仿真结果表明,提出的算法有更高的准确率,同时降低了误检率。

Sample-adaptive-parameters outlier detection method for associated-attributes

In order to solve the interfering problem of associated-attributes in datasets, this paper improved the traditional k-nearest neighbor outlier detection method by the introduction of Mahalanobis distance, and proposed a new sample-based parameters selection method which gained the optimization k-distance value and threshold by training the normal and outlier data in the sample dataset. Simulation results illustrate the proposed algorithm has higher accuracy, lower false detection rate.