一种多源数据驱动的自动交易系统决策模型  

A Multi-source Data Driven Decision Model for Automatic Trading Systems

作　　者：文丹艳 马超群[1] 王琨 WEN Dan-Yan;MA Chao-Qun;WANG Kun(Business School,Hunan University,Changsha 410082)

机构地区：[1]湖南大学工商管理学院,长沙410082

出　　处：《自动化学报》2018年第8期1505-1517,共13页Acta Automatica Sinica

基　　金：国家自然科学基金(71431008;71521061)资助~~

摘　　要：股票自动交易系统属于典型的复杂系统,其成功的关键是如何对股价进行有效的预测与决策.股价受多种信息的影响,但传统的自动交易模型多建立在历史交易数据的基础上.针对上述问题,本文综合利用新闻文本数据与股价技术指标数据,基于人工神经网络(Artificial neural netuorks,ANN)方法设计了一种多源数据驱动的股票自动交易决策模型.本文首先分析了各类财经新闻的特点及其对股价的影响,然后设计了相应模板抽取了中文文本中的财经新闻事件;在此基础上,设计了历史股价和新闻事件数据共同驱动的ANN-News模型,并利用实际数据验证了模型的有效性.实验发现,ANN-News模型比传统的机器学习类模型股价预测准确率提升约4%,收益率提升约7%.Automatic trading systems are typical complex systems, and a successful automatic trading system should be excellent at prediction and decision. Stock prices are affected by the information from various sources, while traditional automatic trading systems only consider the historical trading data. For this issue, we design an automatic trading framework by considering the signals from stock prices and new information based on artificial neural networks. Specifically, we first analyze various kinds of financial events and their corresponding effects on stock prices, and then extract the financial events that have prominent effects on stock prices. Next, we design an automatic trading model driven by stock prices and financial event data. Experiments on real world datasets show that the proposed ANN-News model outperforms the conventional machine learning models by about 4 % in prediction precision and 7 % in return, respectively.

关 键 词：神经网络 数据驱动 自动交易 决策模型 

分 类 号：F832.51[经济管理—金融学]