面向限价指令簿趋势分析的网络集成模型  

Network Ensemble Model for Trend Analysis of Limit Order Books

作　　者：吕雪瑞 张莉[1,2] Lü Xuerui;ZHANG Li(School of Computer Science and Technology,Soochow University,Suzhou 215006;Joint International Research Laboratory of Machine Learning and Neuromorphic Computing,Soochow University,Suzhou 215006)

机构地区：[1]苏州大学计算机科学与技术学院,苏州215006 [2]苏州大学机器学习与类脑计算国际合作联合实验室,苏州215006

出　　处：《模式识别与人工智能》2021年第8期751-759,共9页Pattern Recognition and Artificial Intelligence

基　　金：江苏省高校自然科学研究项目(No.19KJA550002);江苏省六大人才高峰项目(No.XYDXX-054);江苏高校优势学科建设工程项目资助。

摘　　要：为了更好地分析限价指令簿(LOBs)的趋势,文中提出面向LOBs趋势分析的网络集成模型(NEM-LOB).模型融合2个长短期记忆(LSTM)子模型和1个卷积神经网络(CNN)子模型.一个LSTM子模型可通过LOBs的分布信息捕捉全局时间依赖性,另一个LSTM子模型可通过LOBs和订单流的动态信息捕捉全局动态性.CNN子模型通过LOBs的事实信息提取局部特征.最后,结合3个子模型,提取特征以获得预测结果.在FI-2010数据集上的实验表明NEM-LOB通过引入订单流信息,能对LOBs进行更好的趋势分析.To analyze the trend of limit order books(LOBs)better,a network ensemble model for trend analysis of LOBs(NEM-LOB)is proposed.Two long short-term memory(LSTM)sub-models and one convolutional neural network sub-model are integrated in NEM-LOB.One LSTM sub-model captures the global temporal dependence through the distribution information of LOBs.The other LSTM sub-model captures the global dynamics through the dynamic information of LOBs and order streams.The local features are extracted through the factual information of LOBs.Finally,three sub-models are combined to extract features to obtain prediction results.Experiments on FI-2010 dataset show that NEM-LOB makes a better trend analysis for LOBs by combining order streams.

关 键 词：订单流 限价指令簿(LOBs) 集成模型 卷积神经网络(CNN) 长短期记忆(LSTM) 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]