基于词条属性聚类的文本特征选择算法

发布时间：2018-03-16 20:08

本文选题：文本特征选择　切入点：词条属性　出处：《计算机应用研究》2017年02期 　论文类型：期刊论文

【摘要】：文本挖掘之前首先要对文本集进行有效的特征选择。传统的特征选择算法在维数约减及文本表征方面效果有限,并且因需要用到文本的类别信息而不适用于无监督的文本聚类任务。针对这种情况,设计一种适用于文本聚类任务的特征选择算法,提出词条属性的概念。首先基于词频、文档频、词位置及词间关联性构建词条特征模型,重点研究了词位置属性及词间关联性属性的权值计算方法,改进了Apriori算法用于词间关联性属性权值计算;然后通过改进的K-means聚类算法对词条特征模型进行多次聚类完成文本特征选择。实验结果表明,与传统特征选择算法相比,该算法在获得较好维数约减率的同时提高了所选特征词的文本表征能力,能有效适用于文本聚类任务。
[Abstract]:Text mining before the first to carry out the effective feature selection of text set. The traditional feature selection algorithms for dimensionality reduction and text representation and the effect is limited, because of the need to use the category information of the text and is not suitable for the task of unsupervised text clustering. In this case, design a feature selection in text clustering task the algorithm, put forward the concept of entry attribute. Based on word frequency, document frequency, word position and word association between construction characteristic model, focus on the words and word position attribute correlation between attribute weight calculation method, improved Apriori algorithm for word correlation between attribute weight calculation; and then through the improved K-means clustering algorithm multiple entries to cluster feature model to complete the text feature selection. The experimental results show that compared with the traditional feature selection algorithm, this algorithm has better dimension about The reduction rate improves the text representation ability of the selected feature words and can be effectively applied to the text clustering task.

【作者单位】：电子工程学院;
【基金】：国家自然科学基金资助项目(61273302)
【分类号】：TP391.1

【相似文献】