含稀有特征的高平均效用co-location模式挖掘算法  

High Average-Utility Co-Location Patterns Mining Algorithm with Rare Features

作　　者：曾新 李振鹏 杨健 李晓伟 Zeng Xin;Li Zhenpeng;Yang Jian;Li Xiaowei(College of Mathematics and Computer,Dali University,Dali,Yunnan 671003,China)

机构地区：[1]大理大学数学与计算机学院,云南大理671003

出　　处：《大理大学学报》2021年第12期5-11,共7页Journal of Dali University

基　　金：国家自然科学基金项目(71661001,61902049);云南省地方本科高校(部分)基础研究联合专项资金项目(2018FH001-062,2018FH001-063);大理大学数据安全与应用创新团队项目(ZKLX2020308)。

摘　　要：空间高效用co-location模式挖掘以模式中所有特征的参与效用之和为衡量标准,而未考虑模式的长度和稀有特征对模式效用的影响。一般而言,模式的长度越长或存在稀有特征,模式的效用可能越大。在空间高效用co-location模式挖掘研究的基础上,同时考虑模式的长度和可能存在的稀有特征。首先,提出含稀有特征的高平均效用co-location模式挖掘的相关定义;然后,构建含稀有特征的高平均效用co-location模式挖掘算法HAUWR,并在真实和合成数据集下对HAUWR进行大量实验,实验结果表明:HAUWR能够挖掘出满足条件的co-location模式完全集,并具有较好的可扩展性;最后,针对模式长度对高效用co-location模式的影响,HAUWR与含稀有特征的高效用co-location模式挖掘算法HUWR在数据集大小、距离阈值、特征稀有度等方面进行了对比。Spatial high utility co-location patterns mining uses the sum of participating utility of all features in pattern as the measurement standard,without considering the impact of pattern length and rare features.In general,the longer the pattern length or the rarer the features,the greater the utility of the pattern.This paper is based on spatial high-utility co-location pattern mining,while considering the pattern length and possible rare features of patterns.Firstly,the definitions of high average-utility co-location pattern mining with rare features is proposed;then,a high average-utility co-location pattern mining algorithm HAUWR with rare features is constructed,and a large number of experiments on HAUWR are performed on real and synthetic data sets.The experiment result shows that algorithm HAUWR can mine the complete set of co-location patterns that meet the conditions,and has good scalability.Finally,regarding the impact of pattern length of high utility co-location patterns,HAUWR is compared with high utility co-location patterns mining algorithm HUWR containing rare features in some aspects,such as dataset size,distance threshold,and feature rarity.

关 键 词：空间数据挖掘 高平均效用 CO-LOCATION模式 稀有特征 模式长度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]