Prediction method on financial time series data based on matrix profile

doi:10.11772/j.issn.1001-9081.2020060877

Abstract

Abstract: For the fact that institutional trading in the financial market is highly misleading to retail investors in the financial market, a trend prediction method based on the impact of institutional trading behaviors was proposed. First, using the time series Matrix Profile (MP) algorithm and taking the stock turnover rate as the cut-in point, a knowledge base of turnover rate fluctuations based on the influence of institutional trading behaviors under motifs with different lengths was constructed. Second, the motif's length, which leads to the high accuracy of the prediction result of the stock to be predicted was determined. Finally, the fluctuation trend of single stock under the influence of institutional trading behaviors was predicted through the knowledge base of this motif's length. In order to verify the feasibility and accuracy of the new method of trend prediction, the method was compared with Auto-Regressive Moving Average (ARMA) model and Long Short Term Memory (LSTM) network, and the Root-Mean-Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) evaluation indicators were used to compare the 70 stocks' prediction results of three methods. The analysis of experimental results show that, compared with the ARMA model and the LSTM network, in the prediction of 70 stock price trends, the proposed method has more than 80% of the stock prediction results more accurate.

Key words: institutional trading behavior, stock trend prediction, motif discovery, Matrix Profile (MP), time series

摘要： 针对金融市场中机构交易对股票市场中的散户投资行为具有较强的误导性的现象，提出了一种基于机构交易行为影响的趋势预测方法。首先，利用时间序列的矩阵画像（MP）方法，以股票换手率数据为切入点，构建不同兴趣模式长度下的基于机构交易行为影响的换手率波动知识库；其次，确定待预测股票在兴趣模式长度取何值时的预测结果精确度高；最后，根据该兴趣模式长度下的知识库，预测在机构交易行为影响下的单支股票的波动趋势。为验证趋势预测新方法的可行性和准确性，将其与自回归滑动平均（ARMA）模型和长短时记忆（LSTM）网络这两种预测方法进行对比分析，运用均方根误差（RMSE）与平均绝对百分误差（MAPE）评价指标综合比较3种方法对70支股票的预测结果。实验结果分析表明，与ARMA模型和LSTM网络相比，在70支的股票价格趋势预测上，所提方法有80%以上的股票预测结果更准确。

关键词: 机构交易行为, 股票趋势预测, 兴趣模式发现, 矩阵画像, 时间序列

CLC Number:

TP273

GAO Shile, WANG Ying, LI Hailin, WAN Xiaoji. Prediction method on financial time series data based on matrix profile[J]. Journal of Computer Applications, 2021, 41(1): 199-207.

高世乐, 王滢, 李海林, 万校基. 基于矩阵画像的金融时序数据预测方法[J]. 计算机应用, 2021, 41(1): 199-207.

References

[1] 何佳, 何基报, 王霞, 等. 机构投资者一定能够稳定股市吗?——来自中国的经验证据[J]. 管理世界,2007(8):35-42.(HE J,HE J B,WANG X,et al. Can institutional investors be able to stabilize the stock market?-evidence from China[J]. Management World,2007(8):35-42.)
[2] 王咏梅, 王亚平. 机构投资者如何影响市场的信息效率——来自中国的经验证据[J]. 金融研究,2011(10):112-126.(WANG Y M,WANG Y P. How institutional investors affect the information efficiency of markets-evidence from china[J]. Journal of Financial Research,2011(10):112-126.)
[3] 刘京军, 徐浩萍. 机构投资者:长期投资者还是短期机会主义者?[J]. 金融研究,2012(9):141-154.(LIU J J,XU H P. Institutional investors:long-term investors or short-term opportunists?[J]. Journal of Financial Research, 2012(9):141-154.)
[4] 史永东, 王谨乐. 中国机构投资者真的稳定市场了吗?[J]. 经济研究,2014(12):100-112.(SHI Y D,WANG J L. Do Chinese institutional investors really stabilize the market?[J]. Economic Research Journal,2014(12):100-112.)
[5] 王强, 吕政, 王霖青, 等. 基于深度去噪核映射的长期预测模型[J]. 控制与决策,2019,34(5):989-996.(WANG Q,LYU Z, WANG L Q,et al. Deep denoising kernel mapping-based long-term prediction model[J]. Control and Decision, 2019, 34(5):989-996.)
[6] 张贵生, 张信东. 基于微分信息的ARMAD-GARCH股价预测模型[J]. 系统工程理论与实践,2016,36(5):1136-1145. (ZHANG G S,ZHANG X D. A differential-information based ARMAD-GARCH stock price forecasting model[J]. Systems Engineering-Theory and Practice,2016,36(5):1136-1145.)
[7] 吴少聪. 基于混合模型的股票趋势预测方法研究[D]. 哈尔滨:哈尔滨工业大学,2017:61.(WU S C. Research on methods of stock trends prediction based on hybrid model[D]. Harbin:Harbin Institute of Technology,2017:61.)
[8] 宋刚, 张云峰, 包芳勋, 等. 基于粒子群优化LSTM的股票预测模型[J]. 北京航空航天大学学报,2019,45(12):2533-2542. (SONG G,ZHANG Y F,BAO F X,et al. Stock prediction model based on particle swarm optimization LSTM[J]. Journal of Beijing University of Aeronautics and Astronautics,2019,45(12):2533-2542.)
[9] 石浩. 基于递归神经网络的股票趋势预测研究[D]. 北京:北京邮电大学,2018:45.(SHI H. Research on stock trend forecasting based on recurrent neural network[D]. Beijing:Beijing University of Posts and Telecommunications,2018:45.)
[10] 谢琪, 程耕国, 徐旭. 基于神经网络集成学习股票预测模型的研究[J]. 计算机工程与应用,2019,55(8):238-243.(XIE Q, CHENG G G,XU X. Research based on stock predicting model of neural networks ensemble learning[J]. Computer Engineering and Applications,2019,55(8):238-243.)
[11] NAKAGAWA K, IMAMURA M, YOSHIDA K. Stock price prediction using k-medoids clustering with indexing dynamic time warping[J]. Electronics and Communications in Japan,2019,102(2):3-8.
[12] 李海林, 梁叶, 王少春. 时间序列数据挖掘中的动态时间弯曲研究综述[J]. 控制与决策,2018,33(8):1345-1353.(LI H L, LIANG Y,WANG S C. Review on dynamic time warping in time series data mining[J]. Control and Decision,2018,33(8):1345-1353.)
[13] CHEN Y,KEOGH E,HU B,et al. UCR time series classification archive[EB/OL].[2019-09-01]. https://www.cs.ucr.edu/~eamonn/time_series_data/.
[14] YEH C C M,ZHU Y,ULANOVA L,et al. Matrix profile Ⅰ:all pairs similarity joins for time series:a unifying view that includes motifs,discords and shapelets[C]//Proceedings of the IEEE 16th International Conference on Data Mining. Piscataway:IEEE, 2016:1317-1322.
[15] ZHU Y,ZIMMERMAN Z,SENOBARI N S,et al. Matrix profile Ⅱ:Exploiting a novel algorithm and GPUs to break the one hundred million barrier for time series motifs and joins[C]//Proceedings of the IEEE 16th International Conference on Data Mining. Piscataway:IEEE,2016:739-748.
[16] YEH C C M,KAVANTZAS N,KEOGH E. Matrix profile Ⅳ:using weakly labeled time series to predict outcomes[J]. Proceedings of the VLDB Endowment,2017,10(12):1802-1812.
[17] DAU H A,KEOGH E. Matrix profile Ⅴ:a generic technique to incorporate domain knowledge into motif discovery[C]//Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York:ACM,2017:125-134.
[18] YEH C C M,KAVANTZAS N,KEOGH E. Matrix profile Ⅵ:Meaningful multidimensional motif discovery[C]//Proceedings of the 2017 IEEE International Conference on Data Mining. Piscataway:IEEE,2017:565-574.
[19] ZHU Y,IMAMURA M,NIKOVSKI D,et al. Matrix profile Ⅶ:time series chains:a new primitive for time series data mining[C]//Proceedings of the 2017 IEEE International Conference on Data Mining. Piscataway:IEEE,2017:695-704.
[20] ZHU Y,MUEEN A,KEOGH E. Matrix profile Ⅸ:admissible time series motif discovery with missing data[J]. IEEE Transactions on Knowledge and Data Engineering,2019(Early Access):1-1.
[21] ZHU Y,YEH C C M,ZIMMERMAN Z,et al. Matrix profile Ⅺ:SCRIMP++:time series motif discovery at interactive speeds[C]//Proceedings of the 2018 IEEE International Conference on Data Mining. Piscataway:IEEE,2018:837-846.
[22] GHARGHABI S,IMANI S,BAGNALL A,et al. Matrix profile Ⅻ:MPdist:a novel time series distance measure to allow data mining in more challenging scenarios[C]//Proceedings of the 2018 IEEE International Conference on Data Mining. Piscataway:IEEE,2018:965-970.
[23] IMANI S,MADRID F,DING W,et al. Matrix profile XIII:time series snippets:a new primitive for time series data mining[C]//Proceedings of the 2018 IEEE International Conference on Big Knowledge. Piscataway:IEEE,2018:382-389.
[24] YEH C C M,ZHU Y,ULANOVA L,et al. Time series joins, motifs,discords and shapelets:a unifying view that exploits the matrix profile[J]. Data Mining and Knowledge Discovery,2018, 32(1):83-123.series data mining[J]. Control and Decision,2018,33(8):1345-1353.)
[13] CHEN Y,KEOGH E,HU B,et al. UCR time series classification archive[EB/OL].[2019-09-01]. https://www.cs.ucr.edu/~eamonn/time_series_data/.
[14] YEH C C M,ZHU Y,ULANOVA L,et al. Matrix profile Ⅰ:all pairs similarity joins for time series:a unifying view that includes motifs,discords and shapelets[C]//Proceedings of the IEEE 16th International Conference on Data Mining. Piscataway:IEEE, 2016:1317-1322.
[15] ZHU Y,ZIMMERMAN Z,SENOBARI N S,et al. Matrix profile Ⅱ:Exploiting a novel algorithm and GPUs to break the one hundred million barrier for time series motifs and joins[C]//Proceedings of the IEEE 16th International Conference on Data Mining. Piscataway:IEEE,2016:739-748.
[16] YEH C C M,KAVANTZAS N,KEOGH E. Matrix profile Ⅳ:using weakly labeled time series to predict outcomes[J]. Proceedings of the VLDB Endowment,2017,10(12):1802-1812.
[17] DAU H A,KEOGH E. Matrix profile Ⅴ:a generic technique to incorporate domain knowledge into motif discovery[C]//Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York:ACM,2017:125-134.
[18] YEH C C M,KAVANTZAS N,KEOGH E. Matrix profile Ⅵ:Meaningful multidimensional motif discovery[C]//Proceedings of the 2017 IEEE International Conference on Data Mining. Piscataway:IEEE,2017:565-574.
[19] ZHU Y,IMAMURA M,NIKOVSKI D,et al. Matrix profile Ⅶ:time series chains:a new primitive for time series data mining[C]//Proceedings of the 2017 IEEE International Conference on Data Mining. Piscataway:IEEE,2017:695-704.
[20] ZHU Y,MUEEN A,KEOGH E. Matrix profile Ⅸ:admissible time series motif discovery with missing data[J]. IEEE Transactions on Knowledge and Data Engineering,2019(Early Access):1-1.
[21] ZHU Y,YEH C C M,ZIMMERMAN Z,et al. Matrix profile Ⅺ:SCRIMP++:time series motif discovery at interactive speeds[C]//Proceedings of the 2018 IEEE International Conference on Data Mining. Piscataway:IEEE,2018:837-846.
[22] GHARGHABI S,IMANI S,BAGNALL A,et al. Matrix profile Ⅻ:MPdist:a novel time series distance measure to allow data mining in more challenging scenarios[C]//Proceedings of the 2018 IEEE International Conference on Data Mining. Piscataway:IEEE,2018:965-970.
[23] IMANI S,MADRID F,DING W,et al. Matrix profile XIII:time series snippets:a new primitive for time series data mining[C]//Proceedings of the 2018 IEEE International Conference on Big Knowledge. Piscataway:IEEE,2018:382-389.
[24] YEH C C M,ZHU Y,ULANOVA L,et al. Time series joins, motifs,discords and shapelets:a unifying view that exploits the matrix profile[J]. Data Mining and Knowledge Discovery,2018, 32(1):83-123.