基于峰值密度聚类的电信业投诉热点话题检测方法

doi:10.11959/j.issn.1000-0801.2019076

电信科学 ›› 2019, Vol. 35 ›› Issue (5): 97-103.doi: 10.11959/j.issn.1000-0801.2019076

基于峰值密度聚类的电信业投诉热点话题检测方法

江俊^1,^2,³,黄骅⁴,任条娟¹,张登辉¹

¹ 浙江树人大学信息科技学院，浙江杭州 310015
² 浙江大学信息与电子工程学院，浙江杭州 310058
³ 东方通信股份有限公司，浙江杭州 310053
⁴ 万向集团公司万向研究院，浙江杭州 311215

修回日期:2019-04-12 出版日期:2019-05-20 发布日期:2019-05-21
作者简介:江俊（1983- ），男，博士，东方通信股份有限公司工作站及浙江大学流动站博士后，浙江树人大学讲师，主要研究方向为人工智能、数据科学与大数据处理、信息检索等。|黄骅（1983- ），男，博士，浙江大学流动站博士后，主要研究方向为文本挖掘、信息检索和自然语言处理等。|任条娟（1965- ），女，浙江树人大学信息科技学院院长、教授，中国电子学会理事，主要研究方向为无线传感网和车联网、数据科学与大数据处理等。|张登辉（1970- ），男，浙江树人大学信息科技学院副院长、教授，主要研究方向为机器学习与社交计算、数据科学与大数据处理等。

Telecom complaint hot topic detection method based on density peaks clustering

Jun JIANG^1,^2,³,Hua HUANG⁴,Tiaojuan REN¹,Denghui ZHANG¹

¹ College of Information Science and Technology,Zhejiang Shuren University,Hangzhou 310015,China
² College of Information Science and Electronic Engineering,Zhejiang University,Hangzhou 310058,China
³ Eastern Communications Co.,Ltd.,Hangzhou 310053,China
⁴ Wan Xiang Research Institute of Wan Xiang Group Corporation,Hangzhou 311215,China

Revised:2019-04-12 Online:2019-05-20 Published:2019-05-21

摘要/Abstract

摘要：

针对电信业对投诉热点话题缺乏有效的检测方法问题，提出一种基于峰值密度聚类算法的投诉热点话题检测方法。首先建立电信业专用词库用于投诉样本的文本分词，采用向量空间模型表示文本分词，然后通过计算文本分词相似度和密度，并运用密度峰值聚类算法对分词进行聚类分析。最终通过类簇关键词选取并排序，从而得到热点话题描述。将本方法应用到某电信企业投诉热点话题检测中，结果表明本方法有效并具有实际应用价值。

关键词: 热点话题检测, 文本分词, 聚类分析

Abstract:

In view of the lack of effective detection methods for hot topics in telecom industry,a method of complaint hotspots detection based on density peaks clustering algorithm was proposed.Firstly,a special vocabulary for telecommunication industry was established to segment the complaint samples.The vector space model was used to represent the text segmentation.Then,the similarity and density of the text segmentation were calculated,and the clustering analysis of the words was carried out by using the density peaks clustering algorithm.Finally,keywords were selected and sorted by clustering.This method was applied to the complaint hotspots detection tasks within a telecom company.The results show that this method is effective and has practical application value.

Key words: hot topic detection, text segmentation, cluster analysis

中图分类号:

TP391

江俊, 黄骅, 任条娟, 张登辉. 基于峰值密度聚类的电信业投诉热点话题检测方法[J]. 电信科学, 2019, 35(5): 97-103.

Jun JIANG, Hua HUANG, Tiaojuan REN, Denghui ZHANG. Telecom complaint hot topic detection method based on density peaks clustering[J]. Telecommunications Science, 2019, 35(5): 97-103.

图/表 4

图1

表1

表2

表3

参考文献 21

[1]	LI B L , YU S W . Research on topic detection and tracking[J]. Computer Engineering and Applications, 2003,39(17): 7-10.
[2]	关晓惠, 钱亚冠, 孙欣欣 . 一种改进的基于局部密度的聚类算法[J]. 电信科学, 2016,32(1): 54-59.
	GUAN X H , QIAN Y G , SUN X X . An improved clustering algorithm basing on local density[J]. Telecommunications Science, 2016,32(1): 54-59.
[3]	周坚, 石永革, 何美斌 . 基于A-D模型的K-means算法在通话异常客户挖掘中的应用[J]. 电信科学, 2018,34(4): 81-89.
	ZHOU J , SHI Y G , HE M B ,et al. Application of K-means algorithm based on A-D model in calling abnormal customer mining[J]. Telecommunications Science, 2018,34(4): 81-89.
[4]	MOHD M , CRESTANI F , RUTHVEN I . Construction of topics and clusters in topic detection and tracking tasks[C]// International Conference on Semantic Technology and Information Retrieval,June 28-29,2011,Putrajaya,Malaysia. Piscataway:IEEE Press, 2011: 171-174.
[5]	ZHANG X M , LI Z J , CHAO W H . Research of automatic topic detection based on incremental clustering[J]. Journal of Software, 2012,23(6): 1578-1587.
[6]	WANG W , YANG W , QI H F . Netword hotspt topic detection algorithm based on multi-center model[J]. Journal of Nanjing University of Science and Technology, 2009,33(4): 422-421.
[7]	CHENG W , LONG Z Y . Oline topic detection algorithm for internet news[J]. Computer Engineering, 2009,35(18): 28-38.
[8]	LI C , SUN A , DATTA A . Twevent:segment-based event detection from tweets[C]// International Conference on Information and Knowledge Management,Oct 29-Nov 2,2012,San Francisco,USA. Piscataway:IEEE Press, 2012: 155-164.
[9]	VISWANATH P , SURESH B V . Rough-DBSCAN:a fast hybrid density based clustering method for large data sets[J]. Pattern Recognition Letters, 2009,30(16): 1477-1488.
[10]	KUMAR K M , RAMA M R A . A fast DBSCAN clustering algorithm by accelerating neighbor searching using groups method[J]. Pattern Recognition, 2016(58): 39-48.
[11]	谢静瑶, 解思江, 焦阳 ,等. 一种改进的启发式自适应DBSCAN 聚类算法的研究及其在电力系统信息安全预警分析中的应用[J]. 电信科学, 2017,33(Z1): 117-122.
	XIE J Y , XIE S J , JIAO Y ,et al. Research on an improved heuristic auto-adapted DBSCAN clustering algorithm and the application in power system information security early-warning analysis[J]. Telecommunications Science, 2017,33(Z1): 117-122.
[12]	SUN M X , LIU C Q . Research on hot topic detection based on DBSCAN algorithm and inter sentence relationship[J]. Library＆ Information Service, 2017(12): 113-121.
[13]	RODRIGUEZ A , LAIO A . Clustering by fast search and find of density peaks[J]. Science, 2014,344(6191): 1492-1496.
[14]	WEI M , XU W R , JUN G . A chinese text classifier based on N-gram language model and chain augmented na?ve bayesian classifier[J]. Journal of Chinese Information Processing, 2006,20(3): 29-35.
[15]	WU Y , WEI G , LI H . Word segmentation algorithm for chinese language based on N-gram models and machine learning[J]. Journal of Electronics and Information Technology, 2001,23(11): 1148-1153.
[16]	TANG G , XIA Y Q , ZHANG M ,et al. Cross-lingual document clustering based on similarity space model[J]. Journal of Chinese Information Processing, 2016,26(2): 116-120.
[17]	WU C Y . Model of Chinese words rough segmentation based on bi-gram and N-most-probability method[J]. Journal of Computer Applications, 2007,27(2): 2902-2905.
[18]	MA C L , SHAN H , MA T . Improved density peaks based clustering algorithm with strategy choosing cluster center automatically[J]. Computer Science, 2016,43(7): 255-280.
[19]	CAO Z W , ZHOU Y . Design and implementation of JP algorithm based on MapReduce[J]. Computer Engineering, 2012,38(24): 14-16.
[20]	VATHY-FOGARASSY A , KISS A , ABONYI J . Improvement of Jarvis-Patrick clustering based on fuzzy similarity[C]// International Workshop on Fuzzy Logic and Applications:Applications of Fuzzy Sets Theory,July 7-10,2007,Camogli,Italy. New York:ACM Press, 2007: 195-202.
[21]	CHENG Y Y , ZHOU P . Improved K-means clustering algorithm for dynamic allocation cluster center[J]. Computer Technology and Development, 2017,27(2): 33-36.

本文算法	DBSCAN算法		Jarvis-Patrick算法	ADACC K-means算法
λ	minPts	Eps	K	r	M
data1	0.05	5	4.0	5	0.4	10
data2	0.03	6	4.0	6	0.3	20
data3	0.01	6	4.0	7	0.1	30

算法	数据集	准确率	召回率	F值
本文算法	data1	0.667	0.714	0.690
	data2	0.652	0.625	0.639
	data3	0.677	0.719	0.697
DBSCAN算法	data1	0.643	0.529	0.580
	data2	0.484	0.625	0.546
	data3	0.687	0.500	0.579
Jarvis-Patrick算法	data1	0.526	0.714	0.606
	data2	0.609	0.583	0.596
	data3	0.739	0.531	0.618
ADACC K-means算法	data1	0.563	0.643	0.600
	data2	0.636	0.583	0.608
	data3	0.630	0.567	0.597

事件编号	特征词	关键词	热度	对应人工标注的话题名称	对应人工标注的热度
1	用户、要求减免、滞纳、金额、欠费、违约、拆机、帮忙、处理…	用户、要求减免、滞纳金、金额、欠费	0.189	客户要求滞纳金减免	0.081
2	翼支付、账户、周末、活动、消费、未减免、享受、没有…	翼支付、账户、周末、活动、未减免	0.101	翼支付周末活动未减免	0.085
3	手机号码、实名登记、实名、违章停机、复机、核实、号码、小户名、实名制	手机号码、实名登记、实名、违章停机、复机	0.076 2	实名制违章停机复机	0.069 3
4	号码、交费、发票、无法打印、提示、前台、充值、销账、11888…	号码、交费、发票、无法打印、提示	0.050 8	发票无法打印	0.046 2
5	预订单、取消、提示、关联、完成、导购、已完成、提示…	预订单、取消、提示、关联、完成	0.050 6	预订单关联取消	0.030

基于峰值密度聚类的电信业投诉热点话题检测方法

Telecom complaint hot topic detection method based on density peaks clustering

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 4

参考文献 21

相关文章 3

Metrics

推荐阅读 0

[1]	刘子豪, 贾小军, 张素兰, 徐志玲, 张俊. 一种融合MeanShift聚类分析和卷积神经网络的Vibe++背景分割方法[J]. 电信科学, 2021, 37(3): 133-145.
[2]	王炳鑫,侯岩,方红旺,陈雨泽,刘建. 面向“削峰填谷”的电力客户用电行为分析[J]. 电信科学, 2017, 33(5): 164-170.
[3]	翟健宏,李伟,葛瑞海,杨茹. 基于聚类与贝叶斯分类器的网络节点分组算法及评价模型[J]. 电信科学, 2013, 29(2): 51-57.