基于双层堆叠分类模型的水军评论检测

doi:10.6040/j.issn.1671-9352.1.2018.077

《山东大学学报(理学版)》 ›› 2019, Vol. 54 ›› Issue (7): 57-67.doi: 10.6040/j.issn.1671-9352.1.2018.077

基于双层堆叠分类模型的水军评论检测

廖祥文^1,^2,^3,*(),徐阳^1,^2,³,魏晶晶⁴,杨定达^1,^2,³,陈国龙^1,^2,³

1. 福州大学数学与计算机科学学院, 福建福州 350116
2. 福州大学福建省网络计算与智能信息处理重点实验室, 福建福州 350116
3. 数字福建金融大数据研究所, 福建福州 350116
4. 福建江夏学院电子信息科学学院, 福建福州 350108

收稿日期:2018-10-17 出版日期:2019-07-20 发布日期:2019-06-27
通讯作者: 廖祥文 E-mail:liaoxw@fzu.edu.cn
作者简介:廖祥文(1980—),男,博士,副教授,研究方向为信息检索、观点挖掘和情感分析、自然语言处理. E-mail:liaoxw@fzu.edu.cn
基金资助:
国家自然科学基金资助项目(61772135);国家自然科学基金资助项目(U1605251);福建省自然科学基金资助项目(2017J01755);中国科学院网络数据科学与技术重点实验室开放基金课题(CASNDST201708);中国科学院网络数据科学与技术重点实验室开放基金课题(CASNDST201606);北邮可信分布式计算与服务教育部重点实验室主任基金资助(2017KF01)

Review spam detection based on the two-level stacking classification model

Xiang-wen LIAO^1,^2,^3,*(),Yang XU^1,^2,³,Jing-jing WEI⁴,Ding-da YANG^1,^2,³,Guo-long CHEN^1,^2,³

1. College of Mathematics and Computer Science, Fuzhou University, Fuzhou 350116, Fujian, China
2. Fujian Provincial Key Laboratory of Network Computing and Intelligent Information Processing, Fuzhou University, Fuzhou 350116, Fujian, China
3. Digital Fujian Institute of Financial Big Data, Fuzhou 350116, Fujian, China
4. College of Electronics and Information Science, Fujian Jiangxia University, Fuzhou 350108, Fujian, China

Received:2018-10-17 Online:2019-07-20 Published:2019-06-27
Contact: Xiang-wen LIAO E-mail:liaoxw@fzu.edu.cn
Supported by:
国家自然科学基金资助项目(61772135);国家自然科学基金资助项目(U1605251);福建省自然科学基金资助项目(2017J01755);中国科学院网络数据科学与技术重点实验室开放基金课题(CASNDST201708);中国科学院网络数据科学与技术重点实验室开放基金课题(CASNDST201606);北邮可信分布式计算与服务教育部重点实验室主任基金资助(2017KF01)

摘要/Abstract

摘要：

对于水军评论检测问题,已有方法在提取用户行为关系以及通过神经网络提取特征时复杂度过大,同时由于网络评论属于短文本类,其书写的不规范会导致训练过程中文本特征提取困难;另外,已有方法对数据集不平衡分布情况考虑不足。为此,提出了一种基于双层堆叠分类模型的水军评论检测方法。首先通过三元组形式构造矩阵表示用户间关系,并通过主成分分析得到低维用户关系表示,以此刻画用户在评论数据中的行为差异并且降低计算的复杂度;然后,通过评论的段落向量表示以及计算离散型特征(包括文本相似度、信息熵等)解决文本特征难以提取的问题;最后将三者相联结作为融合文本与行为特征的整体特征表示。利用集成学习的方法构造双层堆叠分类模型对评论分类,以提升模型在非平衡数据集下的检测性能。实验采用Yelp2013评论数据集,结果表明,与目前最好的基准方法对比, F₁值提高了1.7%~5.2%,在非平衡数据集中提升尤为明显。

关键词: 水军检测, 特征融合, 集成学习, 主成分分析

Abstract:

For the issue of review spam detection, on the one hand, the time and space complexity of existing methods is high when extracting user behavior relationships and training neural network. On the other hand, the non-standard writing format of E-commercial reviews leads to the indistinct contextual features and most experiment did not consider the effect of the imbalance of data. Therefore, we propose a method for review spam detection based on a two-level stacking classification model. In the method, the relationship between users and products is represented by a triplet. In order to characterize user's behavior and reduce complexity, low-dimensional feature representations are obtained by the principal component analysis. Then, the extracted paragraphs vector representation, information entropy and text similarity is represented as discrete feature to avoid indistinct of contextual features. Finally, the three connections are taken as the overall features combining text and behavioral features. These features are regarded as the input of the two-level stacking classification model in order to improve performance in unbalanced dataset. We conducted experiments in the Yelp 2013 dataset. Experimental results show the F₁ value of our proposed method is 1.7%—5.2% better than the state-of-the-art method. What's more, the classification performance is significantly improved in the unbalanced dataset.

Key words: review detection, feature fusion, ensemble learning, principal component analysis

中图分类号:

TP391

廖祥文,徐阳,魏晶晶,杨定达,陈国龙. 基于双层堆叠分类模型的水军评论检测[J]. 《山东大学学报(理学版)》, 2019, 54(7): 57-67.

Xiang-wen LIAO,Yang XU,Jing-jing WEI,Ding-da YANG,Guo-long CHEN. Review spam detection based on the two-level stacking classification model[J]. JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE), 2019, 54(7): 57-67.

图/表 12

图1

表1

图2

表2

表3

表4

表5

表6

表7

表8

图3

图4

参考文献 17

1	OTT M, CHOI Y, CARDIE C, et al. Finding deceptive opinion spam by any stretch of the imagination[C]// Proceedings of the Meeting of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg: ACM, 2011: 309-319.
2	KIM S, CHANG H, LEE S, et al. Deep semantic frame-based deceptive opinion spam analysis[C]// Proceedings of the 24th ACM International on Conference on Information and Knowledge Management. New York: ACM, 2015: 1131-1140.
3	KO M C, CHEN H H. Analysis of cyber army's behaviours on web forum for elect campaign[C]// Proceedings of the Asia Information Retrieval Symposium. Switzerland: Springer, Cham, 2015: 394-399.
4	LI Huayi, FEI Geli, SHAO Weixiang, et al. Bimodal distribution and co-bursting in review spam detection[C]// Proceedings of the International Conference on World Wide Web. Republic and Canton of Geneva, Switzerland: International World Wide Web Conferences Steering Committee, 2017: 1063-1072.
5	REN Yafeng, ZHANG Yue. Deceptive opinion spam detection using neural network[C]// Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers. Osaka: The COLING 2016 Organizing Committee, 2016: 140-150.
6	WANG Xuepeng, LIU Kang, ZHAO Jun. Handling cold-start problem in review spam detection by jointly embedding texts and behaviors[C]// Proceedings of the Meeting of the Association for Computational Linguistics. Vancouver: ACM, 2017: 366-376.
7	KIM Y. Convolutional neural networks for sentence classification[C]// Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. Doha: EMNLP, 2014: 1746-1751.
8	SANTOSH K C, MAITY S K, MUKHERJEE A. ENWalk: learning network features for spam detection in twitter[C]// Proceedings of the International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation. Switzerland; Springer, Cham, 2017: 90-101.
9	RAYANA S, AKOGLU L. Collective opinion spam detection: bridging review networks and metadata[C]// Proceedings of the 21th ACM Sigkdd International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2015: 985-994.
10	WANG Xuepeng, LIU Kang, HE Shizhu, et al. Learning to represent review with tensor decomposition for spam detection[C]// Proceedings of the Conference on Empirical Methods in Natural Language Processing. Austin: EMNLP, 2016: 866-875.
11	WANG Yalin , SUN Kenan , YUAN Xiaofeng , et al. A novel sliding window PCA-IPF based steady-state detection framework and its industrial application[J]. IEEE Access, 2018, 6: 20995- 21004. doi: 10.1109/ACCESS.2018.2825451
12	LE Q, MIKOLOV T.Distributed representations of sentences and documents[C]// Proceedings of the International Conference on Machine Learning. Beijing: JMLR, 2014: 1188-1196.
13	CHEN Yijun, MAN Leungwong.Optimizing stacking ensemble by an ant colony optimization approach[C]// Proceedings of the 13th Annual Conference Companion on Genetic and Evolutionary Computation. New York: ACM, 2011: 7-8.
14	SANTOSH K C, ARJUN Mukherjee. On the temporal dynamics of opinion spamming: case studies on yelp[C]// Proceedings of the 25th International Conference on World Wide Web. Republic and Canton of Geneva, Switzerland: WWW, 2016: 369-379.
15	MUKHERJEE A, VENKATARAMAN V, LIU B, et al. What yelp fake review filter might be doing[C]// Proceedings of the International AAAI Conference on Web and Social Media. Menlo Park: AAAI, 2013: 409-418.
16	HAI Zeng, ZHAO Peilin, CHENG Peng, et al. Deceptive review spam detection via exploiting task relatedness and unlabeled data[C]// Proceedings of the Conference on Empirical Methods in Natural Language Processing. Austin: EMNLP, 2016: 1817-1826.
17	FAKHRAEI S, SHASHANKA M. Collective spammer detection in evolving multi-relational social networks[C]// Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2015: 1769-1778.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed

类别	特征
用户行为离散特征	用户粉丝数(FC) 用户一天内最大发布评论数(MN) 用户极端评论比例(包括1星或者5星)(ER) 用户评论等级分布信息熵(RE)
评论文本离散特征	该评论与其他同商品评论的平均等级差(AR) 是否是极端评论(1星或5星)(IE) 该评论与同商品评论最大余弦相似度(MC) 该评论被点赞次数(AN)

数据项目	酒店	饭店
水军评论数	802	8 368
非水军评论数	4 876	50 149
水军评论占比	14.1%	14.3%
总评论数	5 678	58 517
总评论者数	5 124	35 593

参数类型	可调参数	参数值
PCA	首次降维目标维度	150
PCA	第二次降维目标维度	(M+N)/2
Doc2vec	段落向量维度	80
Doc2vec	词向量训练模型	c-bow
堆叠分类器	交叉验证次数	5
XGBoost分类器	学习率	0.005
XGBoost分类器	最大深度	5

方法	数据分布	酒店数据集				饭店数据集
方法	数据分布	Precision	Recall	F₁	Accuracy	Precision	Recall	F₁	Accuracy
M_BF+BIGRAM	50:50	82.8	86.9	84.8	85.1	82.8	88.5	85.6	83.3
M_BF+BIGRAM	ND	46.5	82.5	59.4	84.9	48.2	87.9	62.3	78.5
HAAT	50:50	61.3	54.7	57.8	64.4	69.4	59.0	63.8	66.5
HAAT	ND	32.7	53.1	40.8	56.4	35.9	78.9	48.1	68.3
SPEAGLE	50:50	75.7	83.0	79.1	81.0	80.5	83.2	81.8	82.5
SPEAGLE	ND	26.5	56.0	36.0	80.4	48.2	70.5	58.6	82.0
Rescal+BIGRAM	50:50	84.2	89.9	87.0	86.5	86.8	91.8	89.2	89.9
Rescal+BIGRAM	ND	48.2	85.0	61.5	85.9	58.2	90.3	70.8	87.8
Our_Method	50:50	87.3	90.7	88.9	88.8	88.7	93.2	90.9	90.7
Our_Method	ND	52.0	90.0	65.9	86.6	64.6	92.4	76.0	88.3

方法	数据分布	酒店数据集				饭店数据集
方法	数据分布	Precision	Recall	F₁	Accuracy	Precision	Recall	F₁	Accuracy
Rescal	50:50	83.4	87.0	84.5	84.1	82.8	88.5	85.6	83.3
Rescal	ND	44.3	87.7	58.8	83.6	53.9	88.3	66.9	83.4
Our_Method(SVM)	50:50	85.4	86.0	85.7	85.7	86.8	93.1	89.8	89.5
Our_Method(SVM)	ND	48.6	86.3	62.1	85.0	61.3	92.4	73.7	87.3

基于双层堆叠分类模型的水军评论检测

Review spam detection based on the two-level stacking classification model

RichHTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 17

相关文章 7

多维度评价

本文评价

推荐阅读 10

特征	离散型交互特征数据
特征	FC	MN	ER	RE	AR	IE	MC	AN
A	19.0	3.0	0.14	0.79	1.8	0.0	0.02	5.0
B	0.0	1.0	1.0	0.0	2.6	1.0	0.09	0.0

方法	评价指标	酒店数据集 (50:50)	酒店数据集 (ND)	饭店数据集 (50:50)	饭店数据集 (ND)
不考虑离散交互特征	F1	87.2	58.8	90.3	73.1
不考虑离散交互特征	A	86.2	80.1	89.2	86.6
考虑离散交互特征(全特征)	F1	88.9	65.9	90.9	76.0
考虑离散交互特征(全特征)	A	88.8	86.6	90.7	88.3

[1]	李润川,昝红英,申圣亚,毕银龙,张中军. 基于多特征融合的垃圾短信识别[J]. 山东大学学报（理学版）, 2017, 52(7): 73-79.
[2]	于然1,2,刘春阳3*,靳小龙1,王元卓1,程学旗1. 基于多视角特征融合的中文垃圾微博过滤[J]. J4, 2013, 48(11): 53-58.
[3]	邵伟1,祝丽萍2,刘福国2,王秋平2. 对称阵稀疏主成分分析及其在充分降维问题中的应用[J]. J4, 2012, 47(4): 116-120.
[4]	周娟1,王仁卿2,郭卫华2*,王强2,王炜2,庞绪贵3,战金成3,代杰瑞3,周广军4. 鱼台优质稻生产基地土壤地球化学元素调查[J]. J4, 2012, 47(3): 5-9.
[5]	王德良,李科,陆丽玲. 石门国家森林公园唐鱼生境特征分析[J]. J4, 2012, 47(3): 1-4.
[6]	朱世伟,赛英 . 基于主成分分析和粗径向基神经网络的财务预警模型研究[J]. J4, 2008, 43(11): 48-53 .
[7]	杨绍华,林盘,潘晨 . 利用小波变换提高基于KPCA方法的人脸识别性能[J]. J4, 2007, 42(9): 96-100 .