摘要
针对目前数据挖掘中边界点检测效率低、参数阈值范围不容易确定的问题,提出一种新的边界点检测算法BORAL。该算法基于一个有取值范围的参数阈值,利用在边界点的半径ε邻域中边界点与其他点组成的向量夹角中较大的夹角检测边界点,且该夹角邻域内不含有其他点的特征。实验结果表明BORAL能有效检测出边界点、执行效率高,当角度阈值从40°变到57°时,聚类的边界变化不大。
This paper addresses a new boundary detection algorithm called BORAL (Boundary Points Detector Based on Angle), according to the problem of low efficient boundary detection and that it is uneasy to determine the scope of the parameter pruning in data mining. The algorithm based on the parameter pruning with range uses the feature to detect boundary points, such as a bigger angle area of the vectors made of the boundary point with the other points, and the angle area no longer contains any point in ε neighborhood. Experimental results indicate that BORAL detects boundary points effectively and has higher efficiency. The change scope of the border clustering is not large, when angle pruning changes from 40° to 57°.
出处
《计算机工程》
CAS
CSCD
北大核心
2008年第11期203-205,共3页
Computer Engineering
基金
国家自然科学基金资助项目“多语言多媒体网络教学系统的新技术新方法研究”(60363004)
关键词
数据挖掘
边界点
邻域
角度
data mining
boundary points
neighborhood
angle