计算机科学 ›› 2022, Vol. 49 ›› Issue (7): 31-39.doi: 10.11896/jsjkx.210400304

• 数据库&大数据&数据科学* 上一篇    下一篇

基于Bi-LSTM的期货市场关联交易行为检测方法

张源1, 康乐2,3, 宫朝辉3, 张志鸿1   

  1. 1 郑州大学信息工程学院 郑州450001
    2 清华大学计算机科学与技术系 北京100084
    3 郑州易盛信息金融创新实验室 郑州450018
  • 收稿日期:2021-04-29 修回日期:2021-08-16 出版日期:2022-07-15 发布日期:2022-07-12
  • 通讯作者: 张志鸿(iezhzhang@zzu.edu.cn)
  • 作者简介:(iezyzhang@163.com)
  • 基金资助:
    河南省重大公益专项(201300210300)

Related Transaction Behavior Detection in Futures Market Based on Bi-LSTM

ZHANG Yuan1, KANG Le2,3, GONG Zhao-hui3, ZHANG Zhi-hong1   

  1. 1 School of Information Engineering,Zhengzhou University,Zhengzhou 450001,China
    2 Department of Computer Science and Technology,Tsinghua University,Beijing 100084,China
    3 Finance Innovation Laboratory of Zhengzhou Esunny Information,Zhengzhou 450018,China
  • Received:2021-04-29 Revised:2021-08-16 Online:2022-07-15 Published:2022-07-12
  • About author:ZHANG Yuan,born in 1996,postgra-duate,is a student member of China Computer Federation.His main research interests include data mining and financial big data.
    ZHANG Zhi-hong,born in 1965,Ph.D,professor,is a senior member of China Computer Federation.His main research interests include financial big data,blockchain,and distributed computing.
  • Supported by:
    Major Public Welfare Project of Henan Province(201300210300).

摘要: 随着期货市场的不断发展,其交易量屡创新高,但在海量交易的背后,一些交易者利用关联交易行为对市场进行操纵,扰乱了交易秩序,给市场监管和风险控制带来了严峻考验。因此,如何从海量交易中挖掘潜在关联交易行为成为维护期货市场公平交易的重要任务。针对该问题,提出了一种多特征信息融合的双向长短期记忆(Bi-LSTM)网络模型,从原始数据中提取交易时间、交易量、持仓变化、期货品种等多种维度的浅层特征信息,通过Bi-LSTM网络模型从时间序列上向前、向后两个方向的上下文关系学习深层特征,实现关联交易行为检测。针对浅层特征提取提出了一种基于交易行为的多粒度窗口特征提取方法,从日、小时、分钟、秒等级别捕捉账户间交易的关联性,从而解决了原始交易数据维度高、数据量大、关联性弱的问题。模型引入了Dropout策略,缓解了收敛速度慢和过拟合的问题。在郑州商品交易所真实数据上的实验结果表明,与一些传统的分类模型以及RNN和LSTM网络相比,所提方法在分类的准确率和召回率上有明显提升,同时,对特征中各个维度信息的消解实验证明了多特征融合方法和多粒度窗口策略的有效性。另外,抽取了两种期货品种的交易数据进行测试,结果表明所提模型具有良好的泛化能力。

关键词: Bi-LSTM, 多粒度特征提取, 多特征信息融合, 关联交易行为, 期货市场

Abstract: With the continuous development of the futures market,its transaction volume keeps increasing.Behind the massive transactions,some traders use related transaction behaviors to manipulate the futures market and disrupt the order of transactions,which has brought severe challenges to market supervision and risk control.How to mining potential related transaction behaviors from massive transaction data is an important task for maintaining fair transactions in the futures market.In response to this problem,this paper proposes a bidirectional long short-term memory(Bi-LSTM) network model with multi-feature information fusion,which extracts multiple dimensions of shallow feature information such as trading time,trading volume,position changes,and futures varieties from the original transaction data.The Bi-LSTM network model learns deep features from the contextualrelationship in the forward and backward directions of the time series and realizes the detection of related transaction behavior.For shallow features extraction,a multi-granularity window feature extraction method based on transaction behavior is proposed,to captures the correlation of transactions between accounts from the levels of day,hour,minute,second,etc.It solves the problems of high data dimension,large amount of data,and weak correlation of original transaction data.The model introduces the Dropout strategy to alleviate problems of slow convergence and over-fitting.Experimental results on the real data of Zhengzhou Commodity Exchange show that the proposed method evidently improves the classification precision and recall compared with some traditional classification models and RNN and LSTM network.At the same time,the ablation experiment of each dimension information proves the effectiveness of the multi-feature fusion method and the multi-granularity window strategy.In addition,the transaction data of two futures varieties are extracted for testing,and the results show that the proposed model has good generalization ability.

Key words: Bi-LSTM, Futures market, Multi-feature information fusion, Multi-granularity feature extraction, Related transaction behavior

中图分类号: 

  • TP301
[1]HSU C H,LEE H C.Insider trading and information revelation with the introduction of futures markets[J].Economic Mode-lling,2014,43:173-182.
[2]MENG K.The China Securities Regulatory Commission revise the management method of Futures Exchange [EB/OL].(2019-11-16) [2019-11-16].http://epaper.zqrb.cn/html/2019-11/16/content_530969.htm.
[3]CHEN C,ZHAO L,BIAN J,et al.Investment behaviors can tell what inside:Exploring stock intrinsic properties for stock trend prediction[C]//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.New York:ACM,2019:2376-2384.
[4]SZEGEDY C,VANHOUCKE V,IOFFE S,et al.Rethinking the inception architecture for computer vision[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway,NJ:IEEE,2016:2818-2826.
[5]ZHU Z K,LI J B.Multi-Feature Information Fusion LSTM-RNN Detection for OSA[J].Journal of Computer Research and Development,2020,57(12):2547-2555.
[6]SRIVASTAVA N,HINTON G,KRIZHE A,et al.Dropout:A Simple Way to Prevent Neural Networks from Overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958.
[7]FRANKE M,HOSER B.On the analysis of irregular stock market trading behavior[M]//Data Analysis,Machine Learning and Applications.Berlin:Springer,2008:355-362.
[8]PALSHIKAR G K,APTE M.Collusion set detection usinggraph clustering[J].Data Mining & Knowledge Discovery,2008,16(2):135-164.
[9]CAO H B,YU M O,PHILIP S,et al.Detecting abnormal coupled sequences and sequence changes in group-based manipulative trading behaviors[C]//Association for Computing Machi-nery.2010:85-94.
[10]ZHANG X J,ZHANG T L,ZHANG Z H.a clustering detection method of related transaction behavior for futures market[J].Computer Applications and Software,2016(12):275-278.
[11]ZHAI J,CAO Y,YAO Y,et al.Coarse and fine identification of collusive clique in financial market[J].Expert Systems with Applications,2017,69:225-238.
[12]HUANG B,LIU Z,CHEN J,et al.Behavior pattern clustering in blockchain networks[J].Multimedia Tools & Application,2017,76(19):20099-20110.
[13]WANG J,ZHOU S,GUAN J.Detecting potential collusivecliques in futures markets based on trading behaviors from real data[J].Neurocomputing,2012,92:44-53.
[14]YU X P.Related account group identification technology in futures market[D].Shanghai:Shanghai JiaoTong University,2013.
[15]MA X F,FENG Z.identification method of suspected related account based on transaction behavior[J].The Foreland of Trading Technology,2017(12):52-56.
[16]MIAO Z C,SHI G W,LU J D,et al.An Automatic Identification System for Related Behaviors of Financial Derivatives Investors:CN 109636608 A[P].2019-04-16.
[17]HU L.Associated account identification method and electronicequipment:CN 110879816 A[P].2020-03-13.
[18]SUNNY M,MASWOOD M,ALHARBI A G.Deep Learning-Based Stock Price Prediction Using LSTM and Bi-Directional LSTM Model[C]//2020 2nd Novel Intelligent and Leading Emerging Sciences Conference(NILES).2020:87-92.
[19]XIE Y,LIANG R Y,LIANG Z Y,et al.Speech Emotion Classification Using Attention-Based LSTM[J].IEEE/ACM Transactions on Audio,Speech and Language Processing(TASLP),2019,27(11):1675-1685.
[1] 杜琳, 曹东, 林树元, 瞿溢谦, 叶辉.
基于BERT与Bi-LSTM融合注意力机制的中医病历文本的提取与自动分类
Extraction and Automatic Classification of TCM Medical Records Based on Attention Mechanism of BERT and Bi-LSTM
计算机科学, 2020, 47(11A): 416-420. https://doi.org/10.11896/jsjkx.200200020
[2] 冯鸾鸾, 李军辉, 李培峰, 朱巧明.
面向国防科技领域的技术和术语识别方法研究
Technology and Terminology Detection Oriented National Defense Science
计算机科学, 2019, 46(12): 231-236. https://doi.org/10.11896/jsjkx.190300069
[3] 陈俊华, 郝彦惠, 郑丁文, 陈思宇.
基于DBN深度学习的期货市场价格预测建模与决策
Modeling and Decision-making of Futures Market Price Prediction with DBN
计算机科学, 2018, 45(6A): 75-78.
[4] 刘飞龙,郝文宁,陈刚,靳大尉,宋佳星.
基于双线性函数注意力Bi-LSTM模型的机器阅读理解
Attention of Bilinear Function Based Bi-LSTM Model for Machine Reading Comprehension
计算机科学, 2017, 44(Z6): 92-96. https://doi.org/10.11896/j.issn.1002-137X.2017.6A.019
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!