《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (2): 427-436.DOI: 10.11772/j.issn.1001-9081.2025080955
• 人工智能 • 上一篇
魏涵玥, 郭晨娟(
), 梅杰源, 田锦东, 陈鹏, 徐榕荟, 杨彬
收稿日期:2025-08-20
修回日期:2025-09-11
接受日期:2025-10-10
发布日期:2026-03-02
出版日期:2026-02-10
通讯作者:
郭晨娟
作者简介:魏涵玥(1999—),女,江苏苏州人,硕士研究生,主要研究方向:金融时序预测、多模态预测基金资助:
Hanyue WEI, Chenjuan GUO(
), Jieyuan MEI, Jindong TIAN, Peng CHEN, Ronghui XU, Bin YANG
Received:2025-08-20
Revised:2025-09-11
Accepted:2025-10-10
Online:2026-03-02
Published:2026-02-10
Contact:
Chenjuan GUO
About author:WEI Hanyue, born in 1999, M. S. candidate. Her research interests include financial time series forecasting, multimodal forecasting.Supported by:摘要:
现有股票预测模型多基于单一模态,忽视了行业间的联动效应与信息异质性;部分研究虽引入了文本模态,但在处理模态异构所导致的时滞性和多粒度等问题上仍存在不足。因此,提出面向股票市场的融合时频特征与混合文本的多模态股票预测框架MATCH(Multimodal stock prediction frAmework inTegrating time-frequenCy features and Hybrid text)。一方面,设计混合专家(MoE)预训练策略为每个行业构建特定的预训练表征模型,在预测过程中动态选择匹配的专家网络,并注入行业特征信息;另一方面,设计频域分解与层次化融合机制,通过双流预训练架构获取高频未来波动和低频未来趋势的表征,把它们与不同时间尺度的文本信息进行跨模态交互,更精准地捕捉市场动态变化,并实现多粒度场景下的时序与文本有效交互。在2个真实股票数据集S&P 500和CMIN-US上,MATCH与ESTIMATE(Efficient STock Integration with teMporal generative filters and wavelet hypergraph ATtEntions)和PatchTST等主流方法进行对比的实验结果显示,在S&P 500数据集上相较次优基线模型Adv-ALSTM,MATCH的夏普比率(SR)提升了50.5%;在更具有挑战性的CMIN-US数据集上,MATCH的SR提升了2.35%,其余指标均取得了最佳成绩。MATCH预测性能提升明显可为金融多模态数据融合提供新颖且高效的解决方案。
中图分类号:
魏涵玥, 郭晨娟, 梅杰源, 田锦东, 陈鹏, 徐榕荟, 杨彬. 融合时频特征与混合文本的多模态股票预测框架MATCH[J]. 计算机应用, 2026, 46(2): 427-436.
Hanyue WEI, Chenjuan GUO, Jieyuan MEI, Jindong TIAN, Peng CHEN, Ronghui XU, Bin YANG. MATCH: multimodal stock prediction framework integrating time-frequency features and hybrid text[J]. Journal of Computer Applications, 2026, 46(2): 427-436.
| 数据集 | 股票数 | 数据源 | 数据时间区间 | |||
|---|---|---|---|---|---|---|
| 时序 | 文本 | 训练 | 推理 | 测试 | ||
| S&P 500 | 87 | Yahoo Finance | 2014-01-01 to 2014-12-31 | 2015-01-01 to 2015-10-01 | 2015-10-01 to 2015-12-31 | |
| CMIN-US | 110 | Google Finance | Yahoo | 2018-01-02 to 2020-10-15 | 2020-10-16 to 2021-03-11 | 2021-03-12 to 2021-12-31 |
表1 实验数据集的统计信息
Tab. 1 Statistical information of experimental datasets
| 数据集 | 股票数 | 数据源 | 数据时间区间 | |||
|---|---|---|---|---|---|---|
| 时序 | 文本 | 训练 | 推理 | 测试 | ||
| S&P 500 | 87 | Yahoo Finance | 2014-01-01 to 2014-12-31 | 2015-01-01 to 2015-10-01 | 2015-10-01 to 2015-12-31 | |
| CMIN-US | 110 | Google Finance | Yahoo | 2018-01-02 to 2020-10-15 | 2020-10-16 to 2021-03-11 | 2021-03-12 to 2021-12-31 |
| 模型 | S&P 500 | CMIN⁃US | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| IC(↑) | RIC(↑) | ICIR(↑) | RICIR(↑) | SR(↑) | IC(↑) | RIC(↑) | ICIR(↑) | RICIR(↑) | SR(↑) | |
| Autoformer | 0.072 2 | 0.059 7 | 0.788 5 | 0.351 5 | 0.038 7 | 0.717 3 | 0.609 1 | 0.284 3 | ||
| DLinear | 0.059 6 | 0.045 5 | 0.673 0 | 0.537 9 | 0.417 4 | 0.021 5 | 0.039 0 | 0.518 1 | 0.502 8 | 0.344 0 |
| PatchTST | 0.070 7 | 0.054 2 | 0.797 2 | 0.666 8 | 0.745 6 | 0.049 4 | 0.032 9 | 0.608 8 | 0.616 2 | 0.689 1 |
| iTransformer | 0.043 6 | 0.025 7 | 0.545 0 | 0.352 5 | 0.761 9 | 0.035 1 | 0.027 6 | 0.615 5 | 0.505 4 | 0.732 0 |
| Adv-ALSTM | 0.071 4 | 0.036 1 | 0.828 1 | 0.871 8 | 0.066 3 | 0.036 2 | 0.711 6 | |||
| ESTIMATE | 0.115 3 | 0.901 7 | 0.894 3 | 0.036 1 | 0.735 4 | 0.763 3 | ||||
| MATCH | 0.090 1 | 1.151 9 | 1.102 3 | 1.356 0 | 0.081 9 | 0.049 3 | 0.857 7 | 0.831 1 | 0.906 2 | |
表2 在CMIN-US和S&P 500数据集上的实验结果对比
Tab. 2 Comparison of experimental results on CMIN-US and S&P 500 datasets
| 模型 | S&P 500 | CMIN⁃US | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| IC(↑) | RIC(↑) | ICIR(↑) | RICIR(↑) | SR(↑) | IC(↑) | RIC(↑) | ICIR(↑) | RICIR(↑) | SR(↑) | |
| Autoformer | 0.072 2 | 0.059 7 | 0.788 5 | 0.351 5 | 0.038 7 | 0.717 3 | 0.609 1 | 0.284 3 | ||
| DLinear | 0.059 6 | 0.045 5 | 0.673 0 | 0.537 9 | 0.417 4 | 0.021 5 | 0.039 0 | 0.518 1 | 0.502 8 | 0.344 0 |
| PatchTST | 0.070 7 | 0.054 2 | 0.797 2 | 0.666 8 | 0.745 6 | 0.049 4 | 0.032 9 | 0.608 8 | 0.616 2 | 0.689 1 |
| iTransformer | 0.043 6 | 0.025 7 | 0.545 0 | 0.352 5 | 0.761 9 | 0.035 1 | 0.027 6 | 0.615 5 | 0.505 4 | 0.732 0 |
| Adv-ALSTM | 0.071 4 | 0.036 1 | 0.828 1 | 0.871 8 | 0.066 3 | 0.036 2 | 0.711 6 | |||
| ESTIMATE | 0.115 3 | 0.901 7 | 0.894 3 | 0.036 1 | 0.735 4 | 0.763 3 | ||||
| MATCH | 0.090 1 | 1.151 9 | 1.102 3 | 1.356 0 | 0.081 9 | 0.049 3 | 0.857 7 | 0.831 1 | 0.906 2 | |
| 模型 | S&P 500 | CMIN⁃US | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| IC(↑) | RIC(↑) | ICIR(↑) | RICIR(↑) | SR(↑) | IC(↑) | RIC(↑) | ICIR(↑) | RICIR(↑) | SR(↑) | |
| MATCH-v0 | 0.045 4 | 0.026 4 | 0.582 4 | 0.307 7 | 0.908 3 | 0.051 7 | 0.025 8 | 0.601 9 | 0.698 4 | 0.745 7 |
| MATCH-v1 | 0.082 9 | 1.012 5 | 1.003 1 | 0.067 9 | 0.842 2 | 0.758 9 | 0.739 5 | |||
| MATCH-v2 | 0.080 1 | 1.442 2 | 1.247 1 | 0.928 3 | 0.046 7 | 0.862 1 | ||||
| MATCH | 0.090 1 | 0.086 9 | 1.356 0 | 0.081 9 | 0.049 3 | 0.831 1 | 0.906 2 | |||
表3 消融实验结果的对比
Tab. 3 Comparison of ablation experimental results
| 模型 | S&P 500 | CMIN⁃US | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| IC(↑) | RIC(↑) | ICIR(↑) | RICIR(↑) | SR(↑) | IC(↑) | RIC(↑) | ICIR(↑) | RICIR(↑) | SR(↑) | |
| MATCH-v0 | 0.045 4 | 0.026 4 | 0.582 4 | 0.307 7 | 0.908 3 | 0.051 7 | 0.025 8 | 0.601 9 | 0.698 4 | 0.745 7 |
| MATCH-v1 | 0.082 9 | 1.012 5 | 1.003 1 | 0.067 9 | 0.842 2 | 0.758 9 | 0.739 5 | |||
| MATCH-v2 | 0.080 1 | 1.442 2 | 1.247 1 | 0.928 3 | 0.046 7 | 0.862 1 | ||||
| MATCH | 0.090 1 | 0.086 9 | 1.356 0 | 0.081 9 | 0.049 3 | 0.831 1 | 0.906 2 | |||
| [1] | 陈榕,任崇广,王智远,等. 基于注意力机制的CRNN文本分类算法[J]. 计算机工程与设计, 2019, 40(11): 3151-3157. |
| CHEN R, REN C G, WANG Z Y, et al. Attention based CRNN for text classification[J]. Computer Engineering and Design, 2019, 40(11): 3151-3157. | |
| [2] | 王慧斌,胡展傲,胡节,等. 基于分段注意力机制的时间序列预测模型[J]. 计算机应用, 2025, 45(7): 2262-2268. |
| WANG H B, HU Z A, HU J, et al. Time series forecasting model based on segmented attention mechanism[J]. Journal of Computer Applications, 2025, 45(7): 2262-2268. | |
| [3] | 李岚皓,严皓钧,周号益,等. 基于神经网络的多尺度信息融合时间序列长期预测模型[J]. 计算机应用, 2025, 45(6): 1776-1783. |
| LI L H, YAN H J, ZHOU H Y, et al. Multi-scale information fusion time series long-term forecasting model based on neural network[J]. Journal of Computer Applications, 2025, 45(6): 1776-1783. | |
| [4] | 王泉,陆啟想,施珮. 用于交通流量预测的多图扩散注意力网络[J]. 计算机应用, 2025, 45(5): 1472-1479. |
| WANG Q, LU Q X, SHI P. Multi-graph diffusion attention networks for traffic flow prediction[J]. Journal of Computer Applications, 2025, 45(5): 1472-1479. | |
| [5] | BALL R, BROWN P. An empirical evaluation of accounting[J]. Journal of Accounting Research, 1968, 6(2): 159-178. |
| [6] | BAKER S R, BLOOM N, DAVIS S J. Measuring economic policy uncertainty[J]. The Quarterly Journal of Economics, 2016, 131(4): 1593-1636. |
| [7] | PÁSTOR Ľ, VERONESI P. Political uncertainty and risk premia[J]. Journal of Financial Economics, 2013, 110(3): 520-545. |
| [8] | GROSSMAN S J, STIGLITZ J E. On the impossibility of informationally efficient markets[J]. The American Economic Review, 1980, 70(3): 393-408. |
| [9] | SHI B, HSU W N, LAKHOTIA K, et al. Learning audio-visual speech representation by masked multimodal cluster prediction[EB/OL]. [2025-06-10].. |
| [10] | GAVRILYUK K, SANFORD R, JAVAN M, et al. Actor-Transformers for group activity recognition[C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 836-845. |
| [11] | HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. |
| [12] | SALINAS D, FLUNKERT V, GASTHAUS J, et al. DeepAR: probabilistic forecasting with autoregressive recurrent networks[J]. International Journal of Forecasting, 2020, 36(3): 1181-1191. |
| [13] | BAI S, KOLTER J Z, KOLTUN V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[EB/OL]. [2025-06-10].. |
| [14] | SCARSELLI F, GORI M, TSOI A C, et al. The graph neural network model[J]. IEEE Transactions on Neural Networks, 2009, 20(1): 61-80. |
| [15] | ZHOU H, ZHANG S, PENG J, et al. Informer: beyond efficient Transformer for long sequence time-series forecasting[C]// Proceedings of the 35th AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2021: 11106-11115. |
| [16] | WU H, XU J, WANG J, et al. Autoformer: decomposition transformers with auto-correlation for long-term series forecasting[C]// Advances in Neural Information Processing Systems 34. Red Hook: Curran Associates Inc.. 2021: 22419-22430. |
| [17] | ZHOU T, MA Z, WEN Q, et al. FEDformer: frequency enhanced decomposed Transformer for long-term series forecasting[C]// Proceedings of the 39th International Conference on Machine Learning. New York: JMLR.org, 2022: 27268-27286. |
| [18] | CIRSTEA R G, GUO C, YANG B, et al. Triformer: triangular, variable-specific attentions for long sequence multivariate time series forecasting-full version[C]// Proceedings of the 31st International Joint Conference on Artificial Intelligence. California: ijcai.org, 2022: 1994-2001. |
| [19] | ZHANG Y, YAN J. Crossformer: Transformer utilizing cross-dimension dependency for multivariate time series forecasting[EB/OL]. [2025-06-10].. |
| [20] | NIE Y, NGUYEN N H, SINTHONG P, et al. A time series is worth 64 words: long-term forecasting with Transformers[EB/OL]. [2025-06-10].. |
| [21] | ZENG A, CHEN M, ZHANG L, et al. Are Transformers effective for time series forecasting?[C]// Proceedings of the 37th AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2023: 11121-11128. |
| [22] | HU Z, LIU W, BIAN J, et al. Listening to chaotic whispers: a deep learning framework for news-oriented stock trend prediction[C]// Proceedings of the 11th ACM International Conference on Web Search and Data Mining. New York: ACM, 2018: 261-269. |
| [23] | XU Y, COHEN S B. Stock movement prediction from tweets and historical prices[C]// Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Stroudsburg: ACL, 2018: 1970-1979. |
| [24] | LI S, LIAO W, CHEN Y, et al. PEN: prediction-explanation network to forecast stock price movement with better explainability[C]// Proceedings of the 37th AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2023: 5187-5194. |
| [25] | LUO D, LIAO W, LI S, et al. Causality-guided multi- memory interaction network for multivariate stock price movement prediction[C]// Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Stroudsburg: ACL, 2023: 12164-12176. |
| [26] | KOA K J L, MA Y, NG R, et al. Diffusion variational autoencoder for tackling stochasticity in multi-step regression stock price prediction[C]// Proceedings of the 32nd ACM International Conference on Information and Knowledge Management. New York: ACM, 2023: 1087-1096. |
| [27] | WANG H, WANG T, LI S, et al. Heterogeneous interactive snapshot network for review-enhanced stock profiling and recommendation[C]// Proceedings of the 31st International Joint Conference on Artificial Intelligence. California: ijcai.org, 2022: 3962-3969. |
| [28] | XIA H, AO H, LI L, et al. CI-STHPAN: pre-trained attention network for stock selection with channel- independent spatio-temporal hypergraph[C]// Proceedings of the 38th AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2024: 9187-9195. |
| [29] | ZHAO L, KONG S, SHEN Y. DoubleAdapt: a meta- learning approach to incremental learning for stock trend forecasting[C]// Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York: ACM, 2023: 3492-3503. |
| [30] | YOO J, SOUN Y, PARK Y C, et al. Accurate multivariate stock movement prediction via data-axis Transformer with multi-level contexts[C]// Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York: ACM, 2021: 2037-2045. |
| [31] | LI S, SUN Y, LIN Y, et al. CausalStock: deep end- to-end causal discovery for news-driven stock movement prediction[C]// Proceedings of the 38th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2024: 47432-47454. |
| [32] | WU S, İRSOY O, LU S, et al. BloombergGPT: a large language model for finance[EB/OL]. [2025-06-10].. |
| [33] | WANG N, YANG H, WANG C D. FinGPT: instruction tuning benchmark for open-source large language models in financial datasets[EB/OL]. [2025-06-10].. |
| [34] | ZHANG B, YANG H, ZHOU T, et al. Enhancing financial sentiment analysis via retrieval augmented large language models[C]// Proceedings of the 4th ACM International Conference on AI in Finance. New York: ACM, 2023: 349-356. |
| [35] | DONG Z, FAN X, PENG Z. FNSPID: a comprehensive financial news dataset in time series[C]// Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York: ACM, 2024: 4918-4927. |
| [36] | YU X, CHEN Z, LU Y. Harnessing LLMs for temporal data: a study on explainable financial time series forecasting[C]// Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Industry Track. Stroudsburg: ACL, 2023: 739-753. |
| [37] | KOA K J L, MA Y, NG R, et al. Learning to generate explainable stock predictions using self-reflective large language models[C]// Proceedings of the ACM Web Conference 2024. New York: ACM, 2024: 4304-4315. |
| [38] | LI X, SHEN X, ZENG Y, et al. FinReport: explainable stock earnings forecasting via news factor analyzing model[C]// Companion Proceedings of the ACM Web Conference 2024. New York: ACM, 2024: 319-327. |
| [39] | CHEN P, ZHANG Y, CHENG Y, et al. Pathformer: multi- scale transformers with adaptive pathways for time series forecasting[EB/OL]. [2025-06-10].. |
| [40] | HUANG Q, AN Z, ZHUANG N, et al. Harder tasks need more experts: dynamic routing in MoE models[C]// Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Stroudsburg: ACL, 2024: 12883-12895. |
| [41] | RADFORD A, WU J, CHILD R, et al. Language models are unsupervised multitask learners[EB/OL]. [2025-06-10].. |
| [42] | MIKOLOV T, SUTSKEVER I, CHEN K, et al. Distributed representations of words and phrases and their compositionality[C]// Proceedings of the 27th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2013: 3111-3119. |
| [43] | PENNINGTON J, SOCHER R, MANNING C D. GloVe: global vectors for word representation[C]// Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. Stroudsburg: ACL, 2014: 1532-1543. |
| [44] | LIU Y, HU T, ZHANG H, et al. iTransformer: inverted Transformers are effective for time series forecasting[EB/OL]. [2025-06-10].. |
| [45] | FENG F, CHEN H, HE X, et al. Enhancing stock movement prediction with adversarial training[C]// Proceedings of the 28th International Joint Conference on Artificial Intelligence. California: ijcai.org, 2019: 5843-5849. |
| [46] | HUYNH T T, NGUYEN M H, NGUYEN T T, et al. Efficient integration of multi-order dynamics and internal dynamics in stock movement prediction[C]// Proceedings of the 16th ACM International Conference on Web Search and Data Mining. New York: ACM, 2023: 850-858. |
| [1] | 罗虎, 张明书. 基于跨模态注意力机制与对比学习的谣言检测方法[J]. 《计算机应用》唯一官方网站, 2026, 46(2): 361-367. |
| [2] | 王雪, 张丽萍, 闫盛, 李娜, 张学飞. 多模态知识图谱补全方法综述[J]. 《计算机应用》唯一官方网站, 2026, 46(2): 341-353. |
| [3] | 王菲, 陶冶, 刘家旺, 李伟, 秦修功, 张宁. 面向智慧家庭空间的时空知识图谱的双模态融合构建方法[J]. 《计算机应用》唯一官方网站, 2026, 46(1): 52-59. |
| [4] | 吴俊衡, 王晓东, 何启学. 基于统计分布感知与频域双通道融合的时序预测模型[J]. 《计算机应用》唯一官方网站, 2026, 46(1): 113-123. |
| [5] | 黄舒雯, 郭柯宇, 宋翔宇, 韩锋, 孙士杰, 宋焕生. 基于单目图像的多目标三维视觉定位方法[J]. 《计算机应用》唯一官方网站, 2026, 46(1): 207-215. |
| [6] | 李亚男, 郭梦阳, 邓国军, 陈允峰, 任建吉, 原永亮. 基于多模态融合特征的并分支发动机寿命预测方法[J]. 《计算机应用》唯一官方网站, 2026, 46(1): 305-313. |
| [7] | 殷兵, 凌震华, 林垠, 奚昌凤, 刘颖. 兼容缺失模态推理的情感识别方法[J]. 《计算机应用》唯一官方网站, 2025, 45(9): 2764-2772. |
| [8] | 黄锦阳, 崔丰麒, 马长秀, 樊文东, 李萌, 李经宇, 孙晓, 黄林生, 刘志. 基于通用手环的睡眠呼吸暂停检测[J]. 《计算机应用》唯一官方网站, 2025, 45(9): 3045-3056. |
| [9] | 杨青, 朱焱. 改进语言规则中的表示的隐喻识别[J]. 《计算机应用》唯一官方网站, 2025, 45(8): 2491-2496. |
| [10] | 王艺涵, 路翀, 陈忠源. 跨模态文本信息增强的多模态情感分析模型[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2237-2244. |
| [11] | 吴宗航, 张东, 李冠宇. 基于联合自监督学习的多模态融合推荐算法[J]. 《计算机应用》唯一官方网站, 2025, 45(6): 1858-1868. |
| [12] | 张庆, 杨凡, 方宇涵. 基于多模态信息融合的中文拼写纠错算法[J]. 《计算机应用》唯一官方网站, 2025, 45(5): 1528-1534. |
| [13] | 杨雅莉, 黎英, 章育涛, 宋佩华. 面向人脸识别的多模态研究方法综述[J]. 《计算机应用》唯一官方网站, 2025, 45(5): 1645-1657. |
| [14] | 田海燕, 黄赛豪, 张栋, 李寿山. 视觉指导的分词和词性标注[J]. 《计算机应用》唯一官方网站, 2025, 45(5): 1488-1495. |
| [15] | 孙海涛, 林佳瑜, 梁祖红, 郭洁. 结合标签混淆的中文文本分类数据增强技术[J]. 《计算机应用》唯一官方网站, 2025, 45(4): 1113-1119. |
| 阅读次数 | ||||||
|
全文 |
|
|||||
|
摘要 |
|
|||||