深度学习在行业指数技术分析中的应用研究

摘要/Abstract

摘要： 随着全球经济和金融市场的发展，金融资产的价量特征呈现出非线性化演进趋势。考虑到深度学习技术在复杂非线性系统建模方面的优势，越来越多的研究者将深度学习引入到金融分析领域。本文使用深度学习算法分析A股行业指数的技术指标，以提取日频价格序列的变化特征，预测行业指数的涨跌情况，并在A股28个行业指数2000—2017年数据集上进行了实证检验。结果表明，深度学习算法对行业指数的涨跌类型具有一定的预测能力，利用该算法所做出的上涨预测具有较高的可信度，可作为判断指数走势的重要参考；对于不同的行业指数，深度学习算法的预测能力存在差异，上涨预测准确率在55%~65%范围内变动；行业指数在大幅下跌与大幅上涨前夕的技术指标特征具有相似性，深度学习算法对于这两类情形的区分能力较弱。

关键词: 深度学习, 神经网络, 技术分析, 行业指数

Abstract: Investment strategies based on technical analysis have attracted much attention from researchers and investors in A-share market. With the evolution of financial markets, the characteristics of asset price become more and more complicated, and some researchers have pointed out that the performances of classical linear models suffer noticeable degradation in terms of predicting asset price. Considering that deep learning techniques are efficient at capturing intricate quantitative models in nonlinear system, deep-learning-based methods have been introduced into the fields of economic and financial analysis recently. In this paper, deep neutral network (DNN) is applied to analyze the technical indicators and predict the falling and rising tendency of sector indexes in A-share market. The results of empirical analysis indicate that DNN has certain ability in technical analysis. The DNN-based technical analysis method contains four major steps. Firstly, a 36-dimension vector of technical indicators is computed to represent the technical states of sector indexes. Secondly, in order to reduce the dimension of input data and improve the performance of the DNN system, the data of technical indicators are compressed by using two-directional two-dimensional principle component analysis ((2D)²PCA). Thirdly, the DNN system is trained to obtain optimized network coefficients by using historical data. Finally, the test technical features are inputted into DNN to predict the tendency of the sector indexes. The proposed method is validated by using trading data of 28 sector indexes in A-share market between 2000/01/04 and 2017/07/20. Results of empirical test show that the DNN-based method improves the accuracy for predicting the tendency of sector indexes in comparison with methods based on support vector machine (SVM). Furthermore, the prediction accuracy reaches 55%~65% when the DNN-based method is used to distinguish rise and falling tendency, which again implies the potential application value of DNN in technical analysis in A-share markets.

Key words: deep learning, neutral network, technical analysis, sector index

王超. 深度学习在行业指数技术分析中的应用研究[J]. 管理评论, 2021, 33(3): 75-83.

Wang Chao. Study on Application of Deep Learning in Technical Analysis of Sector Indexes[J]. Management Review, 2021, 33(3): 75-83.

参考文献

[1] Malkiel B. G., Fama E. F. Efficient Capital Markets:A Review of Theory and Empirical Work[J]. The Journal of Finance, 1970,25(2):383-417
[2] Pruitt S. W., White R. E. The CRISMA Trading System:Who Says Technical Analysis Can't Beat the Market?[J]. The Journal of Portfolio Management, 1988,14(3):55-58
[3] Brock W., Lakonishok J., LeBaron B. Simple Technical Trading Rules and the Stochastic Properties of Stock Returns[J]. The Journal of Finance, 1992,47(5):1731-1764
[4] Neely C. J., Rapach D. E., Tu J., et al. Forecasting the Equity Risk Premium:The Role of Technical Indicators[J]. Management Science, 2014,60(7):1772-1791
[5] 王志刚. 中国股票市场技术分析有效性研究[D]. 电子科技大学博士学位论文, 2009
[6] 散丹. 市场有效性与技术分析盈利性研究——基于A股市场的实证分析[D]. 西南财经大学硕士学位论文, 2016
[7] 陈曦,庞叶,汪寿阳,等. 一种新的集成预测方法——GPVECM[J]. 系统工程理论与实践, 2008,28(4):108-112
[8] 余乐安,汪寿阳. 基于核主元聚类的股票分类[J]. 系统工程理论与实践, 2009,29(12):1-8
[9] Shah V. H. Machine Learning Techni ques for Stock Prediction[J]. Foundations of Machine Learning, 2007,2007(1):1-19
[10] Patel J., Shah S., Thakkar P., et al. Predicting Stock Market Index Using Fusion of Machine Learning Techniques[J]. Expert Systems with Applications, 2015,42(4):2162-2172
[11] 辛治运,顾明. 基于最小二乘支持向量机的复杂金融时间序列预测[J]. 清华大学学报(自然科学版), 2008,48(7):1147-1149
[12] Yu H., Chen R., Zhang G. A SVM Stock Selection Model within PCA[J]. Procedia Computer Science, 2014,31(1):406-412
[13] 何芳,王小川,张皓. 基于Bootstrap与神经网络模型的浦东新区土地收储增值收益分配研究[J]. 管理评论, 2015,27(12):57-64
[14] 苏治,卢曼,李德轩. 深度学习的金融实证应用:动态、贡献与展望[J]. 金融研究, 2017,(5):111-126
[15] 张承钊. 一种金融市场预测的深度学习模型:FEPA模型[D]. 电子科技大学博士学位论文, 2016
[16] 黄伟伦. 一种股市预测获利行情的深度模型:以台湾股市为例[D]. 台湾大学硕士学位论文, 2017
[17] Singh R., Srivastava S. Stock Prediction Using Deep Learning[J]. Multimedia Tools and Applications, 2017,76(18):18569-18584
[18] 骆双骏. 基于深度学习的黄金期货价格预测[D]. 兰州大学硕士学位论文, 2017
[19] 郑子彬,郑志恒,李阳. 一种结合财经新闻的深度学习股市预测方法[P]. 中国:CN106384166A, 2017
[20] Hinton G. E., Osindero S., Teh Y. W. A Fast Learning Algorithm for Deep Belief Nets[J]. Neural Computation, 2006,18(7):1527-1554
[21] Goodfellow I., Bengio Y., Courville A. Deep Learning[M]. Cambridge:MIT Press, 2016
[22] Shen W., Guo X., Wu C., et al. Forecasting Stock Indices Using Radial Basis Function Neural Networks Optimized by Artificial Fish Swarm Algorithm[J]. Knowledge-Based Systems, 2011,24(3):378-385
[23] Murphy J. J. Technical Analysis of the Financial Markets:A Comprehensive Guide to Trading Methods and Applications[M]. Englewood:Prentice Hall Press, 1999