基于不均匀数据的密度偏差抽样改进算法  

An Improved Alorithm for Density Deviation Sampling Based on Uneven Data

作　　者：吕丹 龙华[1] 高杰[1] 邵玉斌[1] 杜庆治[1] 

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650504

出　　处：《软件导刊》2018年第2期77-79,85,共4页Software Guide

基　　金：2014云南省科技厅基金项目(2014RA051)

摘　　要：针对不均匀数据集的抽样问题,已有随机抽样算法、基于固定网格划分的单维度算法、基于可变网格划分的单维度算法,但仍无法更好地反映数据分布特征问题。在数据挖掘的实际应用中,数据规模越来越大,数据类型也越来越复杂,存在系统整体开销大、时间运行成本高等问题。提出并实现了基于不均匀数据的密度偏差抽样改进算法(IDDS),通过引入网格单元密度和三角函数,从而达到较好的密度偏差抽样效果。实验结果发现,IDDS算法抽样效果更好,提取的样本质量更高,有效保证了不均匀数据的分布特征。与原始的密度偏差抽样算法(DDS)相比,应用IDDS算法的效率更高。Aiming at the sampling problem of inhomogeneous data set,there are random sampling algorithms,single dimension algorithm based on fixed mesh,single dimension algorithm based on variable mesh,but can not better reflect the problem of data distribution.In the practical application of data mining,the scale of the data is increasing and the data type is more and more complicated,and the whole system cost is high,the time running cost is high.In this research,an improved density deviation sampling algorithm（IDDS）based on uneven data was proposed and implemented.According to the introduction of grid cell density and improved trigonometric function,so as to achieve a better density deviation sampling effect.It was found that the improved density deviation sampling algorithm（IDDS）had better sampling effect,the quality of the extracted samples was higher,which ensured the distribution of uneven data effectively.It is concluded that the improved density deviation sampling algorithm（IDDS）results in better efficiency than the original density deviation sampling algorithm（DDS）.This conclusion is important and significant to data mining.

关 键 词：密度偏差抽样算法(DDS) POI信息 数据挖掘 三角函数 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]