基于多Agent传动关系的股市趋势预测

doi:10.19678/j.issn.1000-3428.0067393

摘要/Abstract

摘要：

股市趋势预测是机器学习领域中一个具有挑战性的任务。由于一些因素对于股市的影响是动态且不确定的，导致股市趋势难以预测。针对已有方法在股市预测时存在的灵敏性差、适应力弱等问题，从快变量和慢变量的传动关系出发，利用Agent技术对股市中的快周期和慢周期进行联合建模，提出一种多Agent传动影响图(MATID)股市趋势预测方法。给出股市中快周期和慢周期的划分标准，并引入周期能量的概念；通过对相关趋势指标的特征融合，给出周期能量的量化计算方法；通过分析快周期和慢周期的动态作用过程，给出传动因子的表示方法；将快周期和慢周期分别对应成不同的Agent，利用多Agent影响图模型建模快周期和慢周期的传动过程；利用股市振子模型表示快Agent和慢Agent之间的传动效用，利用联合树的自动推理技术对股市趋势进行预测。在不同样本数量和不同股市趋势下进行实验，结果表明，与门控循环单元、S-LSTM和Hybrid-RNN预测方法相比，MATID方法预测精确率提升1.5%~7.0%，召回率提升5.4%~6.7%，F1值提升3.7%~6.2%，具有良好的灵敏性和适应力。

关键词: 多Agent传动影响图, 周期传动, 振子模型, 效用函数, 联合树

Abstract:

Stock market trend prediction is a challenging problem in the field of machine learning. Owing to the dynamic and uncertain nature of some factors affecting the stock market, changes in stock market trends are difficult to predict. In response to the issues of poor sensitivity and weak adaptability in stock market prediction, starting from the transmission relationship between fast and slow variables, a Multi-Agent Transmission Influence Diagram (MATID) stock market trend prediction method is proposed using Agent technology to jointly model the fast and slow cycles in the stock market. First, the classification criteria of fast and slow periods in the stock market are provided, and the concept of period energy is introduced. Subsequently, a quantitative calculation method for periodic energy through feature fusion of relevant trend indicators is provided; the transmission factor representation method is provided by analyzing the dynamic action process of fast and slow periods. Thereafter, fast and slow periods are corresponded to different Agent, and a multi-Agent influence diagram model is used to model the transmission process of the fast and slow cycles. Furthermore, the stock market oscillator model is used to represent the transmission utility between fast and slow Agent. Finally, the automatic inference technique of a joint tree is used to predict the stock market trend. Experimental results under different sample sizes and different stock market trends demonstrate that compared with Gated Recurrent Unit (GRU), Stacked Long Short-Term Memory (S-LSTM), and Hybrid Recurrent Neural Network (Hybrid-RNN) prediction methods, the MATID method has good sensitivity and adaptability with approximately 1.5%‒7.0%, 5.4%‒6.7%, and 3.7%‒6.2% improvements in prediction, recall, and F1 value, respectively.

Key words: Multi-Agent Transmission Influence Diagram(MATID), periodic transmission, oscillator model, utility function, joint tree

鲍志, 姚宏亮, 方帅, 杨静, 俞奎. 基于多Agent传动关系的股市趋势预测[J]. 计算机工程, 2024, 50(3): 267-276.

Zhi BAO, Hongliang YAO, Shuai FANG, Jing YANG, Kui YU. Prediction of Stock Market Trend Based on Multi-Agent Transmission Relationship[J]. Computer Engineering, 2024, 50(3): 267-276.

http://www.ecice06.com/CN/Y2024/V50/I3/267

图/表 12

图1 上证指数不同周期级别趋势

Fig.1 Trend chart of Shanghai composite index in different cycle levels

图2 MATID模型结构

Fig.2 MATID model structure

图3 1.5时间片MATID联合树

Fig.3 1.5 time slice MATID joint tree

图4 通过接口连接的2个相邻联合树

Fig.4 Two adjacent joint trees connected by an interface

图5 能量调控参数大小对传动效用的影响

Fig.5 Effect of energy regulation parameter size on transmission utility

图6 频率调控参数大小对传动效用的影响

Fig.6 Effect of frequency regulation parameter size on transmission utility

图7 3 033个样本中3种趋势下的预测值与真实值对比

Fig.7 Comparison of predicted values and true values under three trends in 3 033 samples

图8 3种趋势下的2种算法ROC图

Fig.8 ROC charts for two algorithms under three trends

参考文献 26

1	PRAKHAR K, SOUNTHARRAJAN S, SUGANYA E, et al. Effective stock price prediction using time series forecasting[C]//Proceedings of the 6th International Conference on Trends in Electronics and Informatics. Washington D. C., USA: IEEE Press, 2022: 1636-1640.
2	TENGKU N A B T M B, AHMAD K S, AHAD N A, et al. Prediction of FTSE bursa malaysia KLCI stock market using LSTM recurrent neural network[C]//Proceedings of IEEE International Conference on Computing. Washington D. C., USA: IEEE Press, 2022: 415-418.
3	PARK J S, SUNG C H, SUNG L J, et al. Forecasting daily stock trends using random forest optimization[C]//Proceedings of International Conference on Information and Communication Technology Convergence. Washington D. C., USA: IEEE Press, 2019: 1152-1155.
4	PICASSO A, MERELLO S, MA Y K, et al. Technical analysis and sentiment embeddings for market trend prediction. Expert Systems with Applications, 2019, 135, 60- 70. doi: 10.1016/j.eswa.2019.06.014
5	PAVITHYA M B D, PERERA G S D, MUNASINGHE S L, et al. Quantitative analysis and sentiment analysis for stock price forecast: the case of Colombo stock exchange[C]//Proceedings of the 10th International Conference on Information and Automation for Sustainability. Washington D. C., USA: IEEE Press, 2021: 512-517.
6	MEDARHRI I, HOSNI M, NOUISSER N, et al. Predicting stock market price movement using machine learning techniques[C]//Proceedings of the 8th International Conference on Optimization and Applications. Washington D. C., USA: IEEE Press, 2022: 1-5.
7	DINESH S, NITHIN RAO R, ANUSHA S P, et al. Prediction of trends in stock market using moving averages and machine learning[C]//Proceedings of the 6th International Conference for Convergence in Technology. Washington D. C., USA: IEEE Press, 2021: 1-5.
8	WEERATHUNGA H P S D, SILVA A T P. DRNN-ARIMA approach to short-term trend forecasting in forex market[C]//Proceedings of the 18th International Conference on Advances in ICT for Emerging Regions. Washington D. C., USA: IEEE Press, 2018: 287-293.
9	PAN L, MISHRA V. Stock market development and economic growth: empirical evidence from China. Economic Modelling, 2018, 68, 661- 673. doi: 10.1016/j.econmod.2017.07.005
10	XIAO J H, ZHOU M, WEN F M, et al. Asymmetric impacts of oil price uncertainty on Chinese stock returns under different market conditions: evidence from oil volatility index. Energy Economics, 2018, 74, 777- 786. doi: 10.1016/j.eneco.2018.07.026
11	CARDONA L, GUTIÉRREZ M, AGUDELO D A. Volatility transmission between US and Latin American stock markets: testing the decoupling hypothesis. Research in International Business and Finance, 2017, 39, 115- 127. doi: 10.1016/j.ribaf.2016.07.008
12	MORCK R, YEUNG B, YU W. The information content of stock markets: why do emerging markets have synchronous stock price movements?. Journal of Financial Economics, 2000, 58(1/2): 215- 260.
13	WANG B B, XIA X P, XIAO H. Stock price synchronicity and limited arbitrage[C]//Proceedings of International Conference on Management and Service Science. Washington D. C., USA: IEEE Press, 2011: 1-4.
14	WANG L L, WANG Z T, ZHAO S, et al. Stock market trend prediction using dynamical Bayesian factor graph. Expert Systems with Applications, 2015, 42(15/16): 6267- 6275.
15	戚国全, 王浩. 基于影响力传动的Kuramoto股市预测模型. 合肥工业大学学报(自然科学版), 2016, 39(6): 761- 766. doi: 10.3969/j.issn.1003-5060.2016.06.009
	QI G Q, WANG H. Stock market Kuramoto forecasting model based on influence transmission. Journal of Hefei University of Technology (Natural Science), 2016, 39(6): 761- 766. doi: 10.3969/j.issn.1003-5060.2016.06.009
16	MIYANO T, TATSUMI K. Determining anomalous dynamic patterns in price indexes of the London metal exchange by data synchronization. Physica A: Statistical Mechanics and Its Applications, 2012, 391(22): 5500- 5511. doi: 10.1016/j.physa.2012.05.068
17	FUJIMURA K. Centre manifold reduction and the Stuart-Landau equation for fluid motions. Proceedings of the Royal Society of London Series A: Mathematical, Physical and Engineering Sciences, 1997, 453(1956): 181- 203. doi: 10.1098/rspa.1997.0011
18	王朝清. 耦合Stuart-Landau振子系统中的同步相变[D]. 上海: 华东师范大学, 2017.
	WANG Z Q. Synchronous phase transitions in coupled Stuart-Landau oscillator systems [D]. Shanghai: East China Normal University, 2017. (in Chinese)
19	HALE J K, KOÇAK H. Dynamics and bifurcations. Berlin, Germany: Springer, 1991.
20	HOWARD R A, MATHESON J E. Influence diagrams. Decision Analysis, 2005, 2(3): 127- 143.
21	KOLLER D, MILCH B. Multi-agent influence diagrams for representing and solving games. Games and Economic Behavior, 2003, 45(1): 181- 221. doi: 10.1016/S0899-8256(02)00544-4
22	姚宏亮, 王浩, 张佑生, 等. 多Agent动态影响图及其一种近似推理算法研究. 计算机学报, 2009, 31(2): 236- 244. URL
	YAO H L, WANG H, ZHANG Y S, et al. Research on multi-Agent dynamic influence diagrams and its approximate inference algorithm. Chinese Journal of Computers, 2009, 31(2): 236- 244. URL
23	SHEN G Z, TAN Q P, ZHANG H Y, et al. Deep learning with gated recurrent unit networks for financial sequence predictions. Procedia Computer Science, 2018, 131, 895- 903. doi: 10.1016/j.procs.2018.04.298
24	OJO S O, OWOLAWI P A, MPHAHLELE M, et al. Stock market behaviour prediction using stacked LSTM networks[C]//Proceedings of International Multidisciplinary Information Technology and Engineering Conference. Washington D. C., USA: IEEE Press, 2019: 1-5.
25	QIU Y, YANG H Y, LU S, et al. A novel hybrid model based on recurrent neural networks for stock market timing. Soft Computing, 2020, 24(20): 15273- 15290.
26	ADEODATO P, MELO S. Kolmogorov-Smirnov and ROC curve metrics for binary classification performance assessment are equivalent[C]//Proceedings of the 26th International Conference on Pattern Recognition. Washington D. C., USA: IEEE Press, 2022: 1194-1199.

[1]	王志臻,郑烇,陈晨,田洪亮. 基于网络单纯形的虚拟网络映射算法[J]. 计算机工程, 2019, 45(4): 13-17,24.
[2]	刘庆利,姚俊飞,刘治国. AOS中基于跨层效用函数的资源优化方法研究[J]. 计算机工程, 2018, 44(9): 301-308.
[3]	陶洋,王振宇,赵芳金. 异构无线网络中面向多业务的分流策略研究[J]. 计算机工程, 2018, 44(4): 115-119,128.
[4]	李耀华,尚金秋. 基于云计算的飞机PHM体系架构研究[J]. 计算机工程, 2017, 43(12): 6-10.
[5]	宋亚楠，仲茜，曲光亮，李兴立. 多基站协作无线资源分配方法[J]. 计算机工程, 2014, 40(5): 49-53,58.
[6]	叶进，刘建涛，林婧，李陶深. 基于QoS效用函数的比例公平调度算法[J]. 计算机工程, 2014, 40(3): 120-122.
[7]	赵永升. 一种新的(m+1,n)理性秘密分享机制[J]. 计算机工程, 2013, 39(2): 108-111.
[8]	刘华鹏，刘胜全，刘艳，张华楠，李鹏. 基于主体和QoS的语义Web服务组合方法[J]. 计算机工程, 2013, 39(10): 227-231.
[9]	吕翊, 樊思龙, 吴大鹏, 孔晓龙. 带有扩散程度估计的机会网络分组调度机制[J]. 计算机工程, 2012, 38(06): 104-106.
[10]	陈建平, 杨宜民, 魏良. 基于效用值的足球机器人系统任务分配方法[J]. 计算机工程, 2011, 37(01): 197-200.
[11]	水永升, 酆广增. 干扰受限系统的认知用户功率控制算法[J]. 计算机工程, 2010, 36(19): 128-131.
[12]	李志洁;程春田;李慧贤;黄飞雪. 一种基于效用函数的网格资源分配策略[J]. 计算机工程, 2007, 33(24): 1-3.
[13]	冯斌;须文波. 基于粒子群算法的量子谐振子模型[J]. 计算机工程, 2006, 32(20): 18-21.

选择文件类型/文献管理软件名称

选择包含的内容