关键词: 离群点检测, 局部密度, 强k近邻点, 弱k近邻点, 反向k近邻点集

Abstract: Abstract: Mining Outliers is to find exceptional objects that deviates from the most rest of the data set.Outlier detection based on density has attracted lots of attention, existed density-based algorithm LOF is not suitable for data set which is distributed abnormally. Algorithm INFLO solves this problem by analyzing both k nearest neighbor and reverse k nearest neighbor of each data point at the cost of inferior efficiency. Local density-based algorithm LDBO is proposed, which can solve these problems simultaneously and efficiently by adopting vigorous k neighbor point and weak k neighbor point to deal with each data point distinguishingly. Furthermore, to advancing efficiency, prejudgement has applied to data points. Theoretical analysis and experimental results testify that LDBO outperform algorithm INFLO in efficiency, and algorithm LDBO is efficiently and effectively.

Key words: outlier detection, local density, vigorous k neighbor point, weak k neighbor point, reverse k nearest neighbors