一种基于深度学习的时序病变数据段分类方法

doi:10.16180/j.cnki.issn1007-7820.2024.06.011

摘要/Abstract

摘要：

针对在大规模时序医疗数据的分析中现有检测方法检测精度低、检测速度慢等问题,文中提出了一种基于深度学习的时序病变数据段分类方法。该方法在TSTKS(Ternary Search Trees and modified Kolmogorov-Smirnov)算法和滑动窗口理论的基础上,利用深度学习技术实现了对病变数据段的快速准确分类。文中以利用该方法对病变数据段进行分类的结果作为依据,实现了滑动窗口大小的动态调整。通过对真实癫痫脑电信号(Electroencephalogram,EEG)进行分析,证明了所提病变数据段分类方法和基于该分类方法的滑动窗口动态调整机制具有检测速度快、精度较高等优点,可以为大规模时序数据的快速分析研究提供一种新选择。

关键词: 大数据分析, 时序数据, 动态滑动窗口, 多突变点检测, 深度学习, 癫痫脑电信号, BP神经网络, TSTKS算法

Abstract:

In view of the problems of low detection accuracy and slow detection speed in the analysis of large-scale time-series medical data, this study proposes a time-series pathological data segment method based on deep learning. On the basis of TSTKS(Ternary Search Trees and modified Kolmogorov-Smirnov) algorithm and sliding window theory, this method can realize fast and accurate classification of pathological data segments using deep learning technology. Based on the results of classification of pathological data segments by the proposed method, the dynamic adjustment of sliding window size is realized. Through the analysis of real epileptic EEG(Electroencephalogram) signals, the experimental results show that the proposed classification method of pathological data segment and the sliding window dynamic adjustment mechanism based on this classification method have the advantages of fast detection speed and high accuracy, which can provide a new choice for the rapid analysis of large-scale time series data.

Key words: big data analysis, time series data, dynamic sliding window, multiple mutation point detection, deep learning, epileptic EEG, BP neural network, TSTKS algorithm

中图分类号:

TP311

袁傲, 齐金鹏, 贾灿, 薛宇鑫, 郭阳阳. 一种基于深度学习的时序病变数据段分类方法[J]. 电子科技, 2024, 37(6): 84-91.

YUAN Ao, QI Jinpeng, JIA Can, XUE Yuxin, GUO Yangyang. A Classification Method of Time Series Pathological Data Segments Based on Deep Learning[J]. Electronic Science and Technology, 2024, 37(6): 84-91.

图/表 11

图1

图2

图3

图4

图5

图6

图7

表1

表2

图8

表3

参考文献 20

[1]	刘娟. 基于大数据分析的公立医院运营信息化平台建设及应用[J]. 中国总会计师, 2022, 19(11):100-102.
	Liu Juan. Construction and application of public hospital operation information platform based on big data analysis[J]. China Chief Financial Ofiicer, 2022, 19(11):100-102.
[2]	易芳吉, 钟丽莎, 李章勇. 基于SVM分类器的癫痫脑电时空特征提取方法的研究[J]. 重庆邮电大学学报(自然科学版), 2022, 34(3):444-450.
	Yi Fangji, Zhong Lisha, Li Zhangyong. Research on the spatio-temporal feature extraction of epileptic EEG based on SVM classifier[J]. Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition), 2022, 34(3):444-450.
[3]	孔凡书, 齐金鹏, 龚汉鑫, 等. 一种基于波动向量分级技术的病变数据快速分析方法[J]. 电子科技, 2022, 35(7):1-6.
	Kong Fanshu, Qi Jinpeng, Gong Hanxin, et al. A fast analysis method of pathological data based on wave vector grading technology[J]. Electronic Science and Technology, 2022, 35(7):1-6.
[4]	党宁. 基于深度学习多模型决策的癫痫发作类型分类算法研究[D]. 大庆: 东北石油大学,2022:24-27.
	Dang Ning. Research on epileptic seizure type classification algorithm of based on deep learning multi-model decision-making[D]. Daqing: Northeast Petroleum University,2022:24-27.
[5]	贺文韬. 基于人工智能的心脏疾病诊断[J]. 现代商贸工业, 2020, 41(2):188-190.
	He Wentao. Diagnosis of heart diseases based on artificial intelligence[J]. Modern Business Trade Industry, 2020, 41(2):188-190.
[6]	王继成, 吕维雪. 基于符号神经网络的心脏疾病自动诊断[J]. 北京生物医学工程, 1993(3):142-147.
	Wang Jicheng, Lv Weixue. Automatic diagnosis of heart disease based on symbolic neural network[J]. Beijing Biomedical Engineering, 1993(3):142-147.
[7]	邹俊晨, 齐金鹏, 李娜, 等. 一种快速的突变点在线检测算法设计与实现[J]. 电子科技, 2020, 33(8):10-15,52.
	Zou Junchen, Qi Jinpeng, Li Na, et al. Design and implementation of a fast online algorithm for mutation point detection[J]. Electronic Science and Technology, 2020, 33(8):10-15,52.
[8]	Qi J P, Zhang Q, Zhu Y, et al. Anovel method for fast change-point detection on simulated time series and electrocardiogram data[J]. Plos One, 2014, 9(4):1-15.
[9]	Qi J P, Pu F, Zhu Y, et al. Aweighted error distance metrics for performance evaluation on multiple change-point detection in synthetic time series[J]. Computational Intelligence and Neuroscience, 2022(2):1-10.
[10]	曹一彤. 基于动态滑动窗口与模板匹配的病变数据专家系统模型构建[D]. 上海: 东华大学,2022:49-53.
	Cao Yitong. Construction of pathological data expert system model based on dynamic sliding window and template matching[D]. Shanghai: Donghua University,2022:49-53.
[11]	许振伟. BP神经网络分类器在储粮害虫图像检索中的应用研究[J]. 中国粮油学报, 2010, 25(1):103-106.
	Xu Zhenwei. Introducing stored-grain pest image retrieval based on BP neural network[J]. Journal of the Chinese Cereals and Oils Association, 2010, 25(1):103-106.
[12]	周密. 优化BP神经网络在急性胸痛疾病分类诊断中的应用研究[D]. 天津: 天津职业技术师范大学,2020:31-39.
	Zhou Mi. Application of optimized BP neural network in the classification and diagnosis of acute chest pain[D]. Tianjin:Tianjin University of Technology and Education,2020:31-39.
[13]	李娜. 基于突变点检测的癫痫数据异常及其关联网络结构的分析方法[D]. 上海: 东华大学,2020:81-83.
	Li Na. Analysis method of abnormal epilepsy data and its associated network structure based on mutation detection[D]. Shanghai: Donghua University,2020:81-83.
[14]	王伟. 基于BP神经网络的肿瘤数据分类方法研究[D]. 兰州: 兰州交通大学,2017:55-60.
	Wang Wei. The research of classification method of tumor data based on BP neural network[D]. Lanzhou: Lanzhou Jiaotong University,2017:55-60.
[15]	郭芷榕. 基于BP神经网络和半监督学习的时间序列分类模型研究[D]. 太原: 太原理工大学,2018:29-36.
	Guo Zhirong. Research on time series classification model based on BP neural network and semi-supervised learning[D]. Taiyuan: Taiyuan University of Technology,2018:29-36.
[16]	李国平, 张庆伟. 基于BP神经网络的液压挖掘机故障诊断的研究[J]. 机床与液压, 2011, 39(23):160-164.
	Li Guoping, Zhang Qingwei. Research on fault diagnosis for hydraulic excavator based on BP neural network[J]. Machine Tool and Hydraulics, 2011, 39(23):160-164.
[17]	柴绍斌. 基于神经网络的数据分类研究[D]. 大连: 大连理工大学,2007:77-85.
	Chai Shaobin. Research on data classification based on neural network[D]. Dalian: Dalian University of Technology,2007:77-85.
[18]	杨斐, 王坤明, 马欣, 等. 应用BP神经网络分类器识别交通标志[J]. 计算机工程, 2003, 28(10):120-121.
	Yang Fei, Wang Kunming, Ma Xin, et al. Application of BP neural network classifier for road traffic sign recognition[J]. Computer Engineering, 2003, 28(10):120-121.
[19]	王伟平, 李建中, 张冬冬, 等. 数据流上周期更新滑动窗口的连接算法[J]. 哈尔滨工业大学学报, 2005, 5(6):756-759,785.
	Wang Weiping, Li Jianzhong, Zhang Dongdong, et al. Periodically updating sliding window join algorithms overdata streams[J]. Journal of Harbin Institute of Technology, 2005, 5(6):756-759,785.
[20]	李娜, 齐金鹏, 邹俊晨, 等. 一种基于脑电波突变分析的癫痫疾病检测方法[J]. 信息技术, 2019, 43(12):41-46.
	Li Na, Qi Jinpeng, Zou Junchen, et al. An epileptic detection method based on brain wave mutation analysis[J]. Information Technology, 2019, 43(12):41-46.

固定窗口大小	时耗/s	分类准确率/%
2⁶	0.253	90.7
2⁸	0.089	94.4
2¹⁰	0.026	91.7

固定窗口大小	时耗/s	分类准确率/%
2⁶	0.672	89.4
2⁸	0.244	98.8
2¹⁰	0.066	100.0

使用策略	时耗/s	命中率/%
动态滑动窗口	0.119	91.3
窗口长度为2¹¹的固定滑动窗口	0.225	83.2
窗口长度为2¹³的固定滑动窗口	0.057	65.2