«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

30

期数:

2020年08期

页码:

129-133

栏目:

应用开发研究

出版日期:

2020-08-10

文章信息/Info

Title:

Medical Sensitive Text Classification Based on Two-stage Feature Selection

文章编号:

1673-629X(2020)08-0129-05

作者:

陈春玲^*1; 姜慧敏¹; 郭永安²

1. 南京邮电大学 计算机学院、软件学院,江苏 南京 210023; 2. 南京邮电大学 通信与信息工程学院,江苏 南京 210003

Author(s):

CHEN Chun-ling^*1; JIANG Hui-min¹; GUO Yong-an²

1. School of Computer,Nanjing University of Posts and Telecommunications,Nanjing 210023,China; 2. School of Telecommunications & Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China

关键词:

医疗数据; 隐私保护; 特征选择; 敏感数据; 文本分类

Keywords:

medical big data; privacy protection; feature selection; sensitive data; text classification

分类号:

TP18

DOI:

10. 3969 / j. issn. 1673-629X. 2020. 08. 022

摘要:

为完成对医疗数据的敏感性分类,采用文本分类技术从对医疗敏感数据的分类的角度对医疗信息隐私保护进行了研究。 在传统的医疗文本分类基础上,提出基于 LSI-TF-IDF 两阶段特征选择的文本分类方法对医疗文本数据进行敏感性分类。 分别采用基于 TF-IDF 的传统文本分类方法和基于 LSI-TF-IDF 的两阶段特征选择的文本分类方法对糖尿病文本数据进行敏感性分类,利用朴素贝叶斯、KNN、SVM 三个分类器进行实验比较,采用准确率、召回率和 F1值作为评价标准。 实验结果表明,基于 LSI-TF-IDF 两阶段特征选择的文本分类方法较之基于 TF-IDF 的传统文本分类方法在准确率、召回率和 F1值上均有所提升。 证明了该方法在医疗文本数据的敏感性分类上具有更好的分类效果。

Abstract:

In order to complete the sensitive classification of medical data, text classification technology is used to study the privacy protection of medical information from the perspective of classification of medical sensitive data. Based on the traditional medical text classification,a text classification method based on LSI-TF-IDF two-stage feature selection is proposed to classify medical text data. The experiment uses the traditional text class-ification method based on TF-IDF and the text classification method based on LSI-TF-IDF two-stage feature selection to classify the sensitivity of diabetes text data,with three types of naive Bayes,KNN and SVM. For the comparison of experiments,precision ratio, recall ratio and F1 value are used as evaluation criteria. The experiment shows that the text classification method based on LSI-TF-IDF two- stage feature selection has impr-oved in precision ratio,recall ratio and F1 value compared with traditional text classification method based on TF-IDF. It is proved that the proposed method has better classification effect on the sensitivity classification of medical text data.

常用功能

更新日期/Last Update: 2020-08-10