Paper: arXiv 2604.09650

Authors: Xiang Ao, Jingxuan Zhang, Xinyu Zhao

Abstract

Accurately predicting stock repurchases is crucial for quantitative investment and risk management, yet traditional static models fail to capture the complex temporal dependencies of corporate financial conditions. This paper proposes a dynamic early warning system integrating economic theory with deep temporal networks. Using Chinese A-share panel data (2014-2024), we employ a hybrid Temporal Convolutional Network (TCN) and Attention-based LSTM to capture long- and short-term financial evolutionary patterns. Rolling-window cross-validation demonstrates our model significantly outperforms static baselines like Logistic Regression and XGBoost. Furthermore, utilizing Explainable AI (XAI), we reveal the temporal dynamics of repurchase decisions: prolonged “undervaluation” serves as the long-term underlying motive, while a sharp increase in “cash flow” acts as the decisive short-term trigger. This study provides a robust deep learning paradigm for financial forecasting and offers dynamic empirical support for classic corporate finance hypotheses.

Complexity vs Empirical Score

  • Math Complexity: 5.5/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper effectively applies advanced deep temporal neural networks (TCN + Attention LSTM) to corporate financial decision forecasting with solid empirical validation on Chinese A-share data. It bridges machine learning and financial economic theory well using XAI to explain the short- and long-term dynamic drivers of stock repurchases.

Research Flowchart

  flowchart TD
    A[Research Goal: Dynamic Forecasting & Temporal Feature Evolution of Stock Repurchases] --> B[Input Data: Chinese A-Share Panel Data 2014-2024]
    B --> C[Methodology: Hybrid TCN + Attention-LSTM Model]
    C --> D[Computational Process: Rolling-Window Cross-Validation & Baseline Comparison]
    D --> E[Interpretability: Explainable AI Analysis]
    E --> F[Key Outcome 1: TCN-Attention-LSTM Outperforms Static Models]
    E --> G[Key Outcome 2: Long-Term Undervaluation Motive + Short-Term Cash Flow Trigger]

Cited References (47)

  1. The cash flow sensitivity of cash (2004)
  2. Market timing and capital structure (2002)
  3. Payout policy in the 21st century (2005)
  4. Common stock repurchases: An analysis of returns to bondholders and stockholders (1981)
  5. Why do firms repurchase stock (2000)
  6. Do measures of financial constraints measure financial constraints? (2016)
  7. Dividends, share repurchases, and the substitution hypothesis (2002)
  8. Market underreaction to open market share repurchases (1995)
and 39 more
  1. Agency costs of free cash flow, corporate finance, and takeovers (1986)
  2. Estimating standard errors in finance panel data sets: Comparing approaches (2008)
  3. Actual share reacquisitions in open-market repurchase programs (1998)
  4. Common stock repurchases and market signalling: An empirical study (1981)
  5. Machine learning in business and finance: a literature review and research opportunities (2024)
  6. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling (2018) — link
  7. Neural machine translation by jointly learning to align and translate (2014) — link
  8. Random forests (2001)
  9. Xgboost: A scalable tree boosting system (2016)
  10. Support-vector networks (1995)
  11. The relationship between Precision-Recall and ROC curves (2006)
  12. An introduction to ROC analysis (2006)
  13. Deep learning with long short-term memory networks for financial market predictions (2018)
  14. Deep learning (2015)
  15. Empirical asset pricing via machine learning (2020)
  16. Learning from imbalanced data (2009)
  17. Long short-term memory (2012)
  18. Leakage in data mining: Formulation, detection, and avoidance (2012)
  19. Deep neural networks, gradient-boosted trees, random forests: Statistical arbitrage on the S&P 500 (2017)
  20. Enhancing the locality and breaking the memory bottleneck of transformer on time series forecasting (2019)
  21. Temporal fusion transformers for interpretable multi-horizon time series forecasting (2021)
  22. Focal loss for dense object detection (2017)
  23. Wavenet: A generative model for raw audio (2016) — link
  24. Predicting stock and stock price index movement using trend deterministic data preparation and machine learning techniques (2015)
  25. Financial time series forecasting with deep learning: A systematic literature review: 2005–2019 (2020)
  26. Universal features of price formation in financial markets: perspectives from deep learning (2021)
  27. Combining multiple feature selection methods for stock prediction: Union, intersection, and multi-intersection approaches (2010)
  28. Attention is all you need (2017)
  29. Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI (2020)
  30. Machine learning explainability in finance: an application to default risk analysis (2019)
  31. Explainable machine learning in credit risk management (2021)
  32. A unified approach to interpreting model predictions (2017)
  33. From local explanations to global understanding with explainable AI for trees (2020)
  34. Research on the Stock Price Prediction Model Based on Large-Scale Transactions (2025) — link
  35. A hybrid LSTM-GRU model for stock price prediction (2025)
  36. Explainable ai in finance: Addressing the needs of diverse stakeholders (2025)
  37. Explainable artificial intelligence (XAI) in finance: a systematic literature review (2024)
  38. The Application of Artificial Intelligence to Stock Forecasting: A Literature Review (2025)
  39. Stock prediction using machine learning a review paper (2017)

0 of 47 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.