针对股票市场变化快、关系复杂的情况，提出一种结合个股新闻的股票价格预测方法。首先通过动态时间规整算法找到与目标个股序列相似度最高的基准序列，然后通过平滑突变点模型提取新闻影响的长度和时间，转化为时间序列数据，通过统计模型将股票之间的关系引入到时间序列预测，考察新闻影响力与股票历史价格数据之间的关系，同时利用长短期记忆网络将新闻影响与个股数据结合起来进行价格预测。结果表明，新闻在科技类行业领域的股票板块影响力最为明显；相比于已有的股票预测方法，融合模型的预测性能有所提升，并且随时间增长预测精度下降幅度较小。融合模型可以更精确地描述股票价格的变化，在模拟投资策略的条件下取得了14.50%的平均收益。
Aiming at the rapid changes and complex relations in the stock market, this paper proposes a stock price prediction method based on individual stock news. First, through dynamic time warping algorithm，the benchmark sequence with the highest similarity to the target individual stock sequence is found, and then we can extract the length and time of news impact through smooth-and-abrupt change point model, which is converted into time series data. We introduce the relationship between stocks into time series forecasting through statistical models, examine the relationship between news influence and historical stock price data, and combine news influence with individual stock data for price forecasting by using long-and-short-term memory network. The results show that the stock sector’s influence of news in the technology sector is the most obvious. Compared to existing stock prediction methods, the prediction performance of the fusion model has been improved, and the prediction accuracy has decreased slightly over time. The fusion model can more accurately describe the changes in stock prices, achieving an average return of 14.50% under the conditions of simulating investment strategies.
[1] Farias Nazário R T, e Silva J L, Sobreiro V A, et al. A literature review of technical analysis on stock markets[J]. The Quarterly Review of Economics and Finance, 2017, 66: 115-126. DOI: 10.1016/j.qref. 2017.01.014.
[2] 徐飏, 王艳, 赵子龙. 基于资产选择的投资组合模型与中国股市实证检验[J]. 中国科学院大学学报, 2015, 32(4): 446-452. DOI: 10.7523/j.issn. 2095-6134. 2015. 04.004.
[3] 邓凯旭, 宋宝瑞. 小波变换在金融数据分析中的应用[J]. 数理统计与管理, 2006, 25(2): 215-219. DOI: 10.13860/j.cnki.sltj.2006.02.016.
[4] Xu Y, Zhang G S. Application of Kalman filter in the prediction of stock price[C]//Proceedings of the 5th International Symposium on Knowledge Acquisition and Modeling. June 27-28, 2015. London, UK. Paris, France: Atlantis Press, 2015: 197-198. DOI: 10.2991/kam-15.2015.53.
[5] Ariyo A A, Adewumi A O, Ayo C K. Stock price prediction using the ARIMA model [C]//2014 UKSim-AMSS 16th International Conference on Computer Modelling and Simulation. March 26-28, 2014, Cambridge, UK. IEEE, 2015: 106-112. DOI: 10.1109/UKSim.2014.67.
[6] 叶五一, 赵晋海, 缪柏其. 欧美与国内股市流动性风险间的相互关系及风险溢出效应研究：流行病爆发背景下的分析[J]. 数理统计与管理, 2021, 40(2): 292-309. DOI: 10.13860/j.cnki.sltj. 20201205-002.
[7] 次必聪，张品一. 基于ARIMA-LSTM模型的金融时间序列预测[J].统计与决策, 2022, 38(11): 145-149. DOI: 10.13546/j.cnki.tjyjc.2022.11.029.
[8] Feng F L, He X N, Wang X, et al. Temporal relational ranking for stock prediction[J]. ACM Transactions on Information Systems, 37(2)Article No. 27, DOI: 10.1145/3309547.
[9] Chen C, Zhao L, Bian J, et al. Investment behaviors can tell what inside: exploring stock intrinsic properties for stock trend prediction [C]//Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. August 4-8, 2019, Anchorage, AK, USA. New York: ACM, 2019: 2376-2384. DOI: 10.1145/3292500. 3330663.
[10] Tetlock P C, Saar-Tsechansky M, Macskassy S. More than words: quantifying language to measure firms’ fundamentals[J]. The Journal of Finance, 2008,63(3):1437-1467. DOI: 10.1111/j.1540-6261. 2008.01362.x.
[11] 段江娇,刘红忠,曾剑平.投资者情绪指数、分析师推荐指数与股指收益率的影响研究：基于我国东方财富网股吧论坛、新浪网分析师个股评级数据[J].上海金融,2014(11):60-64.DOI:10.13910/j.cnki.shjr.2014.11.012.
[12] Ewing B T. The transmission of shocks among S&P indexes[J]. Applied Financial Economics, 2002, 12(4):285-290. DOI: 10.1080/09603100110090172.
[13] 杜伟锦, 何桃富. 我国证券市场的板块联动效应及模糊聚类分析[J]. 商业研究, 2005(22): 41-45. DOI: 10.13902/j.cnki.syyj.2005.22.014.
[14] Salvador S, Chan P. Toward accurate dynamic time warping in linear time and space[J]. Intelligent Data Analysis, 2007, 11(5):561-580. DOI:10.3233/IDA-2007-11508.
[15] Chen J, Gupta A K. A Bayesian approach to the statistical analysis of a smooth-abrupt change point model[J]. Advances and Applications in Statistics, 2007,7(1):115-125.
[16] Graves A. Generating sequences with recurrent neural networks[EB/OL]. 2013. arXiv: 1308.0850. (2014-06-05)[2023-02-29].https://arxiv.org/abs/1308.0850.
[17] Yan R, Song Y P, Wu H A. Learning to respond with deep neural networks for retrieval-based human-computer conversation system[C]// Proceedings of the 39th International ACM SIGIR conference on Research and Development in Information Retrieval. July 17-21, 2016, Pisa, Italy. New York, NY, USA: ACM, 2016: 55-64. DOI: 10.1145/2911451. 2911542.
[18] Chen J Y, Li S E, Tomizuka M. Interpretable end-to-end urban autonomous driving with latent deep reinforcement learning[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(6): 5068-5078. DOI: 10.1109/TITS.2020.3046646.
[19] Mihalcea R, Tarau P. TextRank: bringing order into Text[C/OL]//Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing. Barcelona, Spain: Association for Computational Linguistics,2004:404-411(2004-07-01)[2023-02-29]. https://web.eecs.umich.edu/~mihalcea/papers/mihalcea.emnlp04.pdf.
[20] Jiang F W, Lee J, Martin X, et al. Manager sentiment and stock returns[J]. Journal of Financial Economics, 2019, 132(1): 126-149. DOI: 10.1016/j.jfineco.2018. 10.001.
[21] Campbell J Y, Thompson S B. Predicting excess stock returns out of sample: can anything beat the historical average?[J]. The Review of Financial Studies, 2008, 21(4): 1509-1531. DOI: 10.1093/rfs/hhm055.
