农晓锋


摘 要: 提出利用基于多目标优化软子空间聚类理论的关联规则数据挖掘方法对高维数据集中局部离散文本数据实现数据特征有效挖掘。首先,利用多目标优化软子空间聚类思想结合非支配排序遗传理论优化加权类内紧致及加权类间分离函数,获取优化后的目标函数及非占优Pareto最优解集,运用加权子空间划分方法对最优解集完成特征聚类;其次,基于关联规则思想运用一种特征提取和关联文本的识别方法,对聚类后的文本特征进行文本间及文本内部的特征识别和分类,即实现了文本信息数据的有效挖掘。实验证明,利用多目标优化软子空间聚类数据挖掘方法可以有效实现高维集中局部离散文本数据的挖掘。
关键词: 高维数据; 数据特征聚类; 数据挖掘; 关联规则
中图分类号: TN911.1?34; TP391 文献标识码: A 文章编号: 1004?373X(2017)19?0138?04
Research on local discrete text data mining method in high?dimensional dataset
NONG Xiaofeng
(Modern Educational and Technological Center, Guilin Tourism University, Guilin 541006, China)
Abstract: An association rules data mining method based on the theory of multi?objective optimization soft subspace clustering is proposed to mine the data feature of local discrete text data in high?dimensional dataset effectively. The thought of multi?objective optimization soft subspace clustering is combined with the theory of non?dominated sorting genetic optimization to optimize the weighted intra?class compactness and weighted inter?class separation function, and obtain the optimized objective function and non?dominated Pareto optimal solution set. The weighting subspace classification method is used to cluster the features of the optimal solution set. A recognition method for feature extraction and text association based on the thought of association rules is used to recognize and classify the features among texts and within texts for the clustered text features, which can realize the effective mining of the text information data. The experimental results show that the data mining method of multi?objective optimization soft subspace clustering can realize the local discrete text data mining in high?dimensional dataset effectively.
Keywords: high?dimensional data; data feature clustering; data mining; association rule
0 引 言
文献[1]指出,在人工智能和数据库领域中,目前各种数据挖掘方法也获得了不同程度的关注。20世纪末开始,人们对各种不同的数据挖掘方法进行深入研究。数据挖掘作为一种决策支持手段,帮助各个领域的专家和开发人员分析各种类型的数据[2?3],然后从中挖掘出潜在的模式并做出正确决策判断。文献[4]中提到数据挖掘通常会利用人工智能、机器学习、模式识别、统计学、可视化等技术来实现该过程。
当前数据挖掘研究领域发展迅速,其面临的问题与挑战也越来越多。第一,越来越大的数据规模,也称之为大规模数据问题;……