基于相对密度的孤立点和边界点识别算法

Recognition Algorithm of Outlier and Boundary Points Based on Relative Density

根据孤立点是数据集合中与大多数数据的属性不一致的数据,边界点是位于不同密度数据区域边缘的数据对象,提出了基于相对密度的孤立点和边界点识别算法(OBRD)。该算法判断一个数据点是否为边界点或孤立点的方法是:将以该数据点为中心、r为半径的邻域按维平分为2个半邻域,由这些半邻域与原邻域的相对密度确定该数据点的孤立度和边界度,再结合阈值作出判断。实验结果表明,该算法能精准有效地对多密度数据集的孤立点和聚类边界点进行识别。

According to the fact that outlier points are the data that are inconsistent with most of data in a data set,and that boundary points are located on the edge of data area with different densities,an algorithm based on relative density was proposed to determine the outlier and boundary points.Through dividing the neighborhood area,which is centered by this point with a radius of r,into two semi-neighborhood areas,and determining this data point's isolation level and boundary level based on the relative density of these semi-neighborhood areas with the original neighborhood area,a final judgment whether a data point is boundary or outlier point can be made according to the threshold value.Experimental results indicate that this algorithm can effectively and accurately identify the outlier and boundary points from multidensity data sets.