基于密度的孤立点检测算法改进研究
Improvement of Detection Algorithm Density-Based Local Outlier
吕奔 1高茂庭1
作者信息
- 1. 上海海事大学信息工程学院，上海 201306
- 折叠
摘要
针对基于密度的孤立点检测算法LOF时间复杂度高的问题，通过优化数据对象邻域查询过程，提出一种两阶段的改进算法DBLOF，先采用DBSCAN聚类算法对数据集进行预处理，去除大部分的非孤立点，得到可能异常数据集，然后再利用LOF算法计算可能异常数据集中对象的局部异常因子并以此找出真正的孤立点。实验结果表明，改进算法能实现有效的局部孤立点检测，并能够降低算法时间复杂度。
Abstract
For the high time complexity of the density-based outlier detecting algorithm (LOF algorithms), proposes an improved algorithm DBLOF with two-stage by optimizing the neighborhood query operation of adjacent objects for each data object. Firstly, clustering algorithm DB-SCAN is taken to preprocess the dataset and remove the most of the non-outliers data objects to get the dataset of all possible outliers. Then, the local outlier factors are calculated on the possible outliers dataset for each data object to find out the real outliers. The experi-ments demonstrate that the proposed algorithm can realize the effective local outlier detection and reduce the time complexity of the algo-rithm.
关键词
孤立点检测/LOF/DBSCAN/聚类/数据挖掘Key words
Outlier Detection/LOF/DBSCAN/Clustering/Data Mining引用本文复制引用
基金项目
国家自然科学基金(61202022)
上海海事大学科研项目()
出版年
2015
NETL
NSTL
维普
万方数据