本研究結合稀疏正交因子迴歸(SOFAR) 與機器學習方法，探討基本面潛在結構與市場價量訊號在台股報酬預測中之資訊互補性。樣本為台灣上市上櫃公司（不含TDR），樣本期間 2005 至 2025 年，樣本外回測期間 2016 至 2025 年，共 119 個月。本研究採用 SOFAR 從 25 個季頻財報變數中萃取稀疏弱因子，結合邊緣分配法、硬閾值除躁與符號校正機制，建構三種特徵集合搭配三種演算法共九個預測模型，並以 OLS-3 為線性基準，合計十個模型。實證結果顯示，Hybrid 特徵集在三種演算法下均取得最高夏普比率，XGBoost-Hybrid 表現最佳，扣除交易成本後年化報酬約13.61%、夏普比率1.425，Newey-West t 統計量達5.541，連續九年維持正夏普比率；特徵重要性分析顯示兩類訊號貢獻相近（市場面44.6%、SOFAR 因子55.4%），跨越檢定截距項在1% 水準下顯著，確認Hybrid 模型產生無法被單一特徵集複製之超額報酬。

This study combines Sparse Orthogonal Factor Regression (SOFAR) with machine learning methods to investigate the informational complementarity between latent fundamental structure and market price-volume signals in Taiwan stock return prediction. The sample consists of Taiwan listed and OTC common stocks (excluding TDRs), covering 2005 to 2025, with an out-of-sample backtesting period from 2016 to 2025 totaling 119 months. SOFAR is employed to extract sparse weak factors from 25 quarterly financial statement variables, combined with the Edge Distribution method, hard-thresholding denoising, and sign correction mechanisms, constructing three feature sets paired with three algorithms yielding nine predictive models, with OLS-3 serving as the linear benchmark for a total of ten models. Empirical results show that the Hybrid feature set achieves the highest Sharpe ratio across all three algorithms. XGBoost-Hybrid performs best with an annualized net return of approximately 13.61%, a Sharpe ratio of 1.425, a Newey-West t-statistic of 5.541, and positive Sharpe ratios for nine consecutive years. Feature importance analysis reveals comparable contributions from both signal types (market features: 44.6%. SOFAR factors: 55.4%), and the spanning test intercept is significant at the 1% level, confirming that the Hybrid model generates excess returns unreplicable by any single feature set.

王芯儀，徐政義，陳姿伶與賴弘能(2024)，「因子訂價模型有效性之比較：臺灣股市實證」，《證券市場發展季刊》，36(2)，1–64。
陳力慈(2023)，《因子動能策略在台灣股票市場的實證研究》，國立臺灣大學財務金融學系碩士論文。
臺灣證券交易所(2026 年1 月13 日)，「2025 年12 月暨第四季集中市場相關數據」，TWSE 臺灣證券交易所。https://accessibility.twse.com.tw/zh/about/news/news/content.html?8a8216d69a3d6cf9019bb6ea2eaf071f
Breiman, L. (2001), “Random Forests,” Machine Learning, 45(1), 5–32.
Bryzgalova, S., Lerner, S., Lettau, M. and Pelger, M. (2025), “Missing Financial Data,” The Review of Financial Studies, 38(3), 803–882.
Bui, D. G., Kong, D.-R., Lin, C.-Y. and Lin, T.-C. (2023), “Momentum in Machine Learning:Evidence from the Taiwan Stock Market,” Pacific-Basin Finance Journal, 82, Article 102178.
Carhart, M. M. (1997), “On Persistence in Mutual Fund Performance,” The Journal of Finance,52(1), 57–82.
Chen, T. and Guestrin, C. (2016), “XGBoost: A Scalable Tree Boosting System,” Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794.
Chen, K. and Wang, W. (2025), rrpack: Reduced-Rank Regression (Version 0.1-14) [Computer software], https://CRAN.R-project.org/package=rrpack.
Cochrane, J. H. (2011), “Presidential Address: Discount Rates,” The Journal of Finance, 66(4),1047–1108.
Fama, E. F. and French, K. R. (1993), “Common Risk Factors in the Returns on Stocks and Bonds,” Journal of Financial Economics, 33(1), 3–56.
Fama, E. F. and French, K. R. (2015), “A Five-Factor Asset Pricing Model,” Journal of Financial Economics, 116(1), 1–22.
Friedman, J. H. (2001), “Greedy Function Approximation: A Gradient Boosting Machine,”Annals of Statistics, 29(5), 1189–1232.
Gibbons, M. R., Ross, S. A. and Shanken, J. (1989), “A Test of the Efficiency of a Given Portfolio,” Econometrica, 57(5), 1121–1152.
Green, J., Hand, J. R. M. and Zhang, X. F. (2017), “The Characteristics that Provide Independent Information about Average U.S. Monthly Stock Returns,” The Review of Financial Studies,30(12), 4389–4436.
Gu, S., Kelly, B. and Xiu, D. (2020), “Empirical Asset Pricing via Machine Learning,” The Review of Financial Studies, 33(5), 2223–2273.
Harvey, C. R., Liu, Y. and Zhu, H. (2016), “…and the Cross-Section of Expected Returns,” The Review of Financial Studies, 29(1), 5–68.
Hastie, T., Tibshirani, R. and Friedman, J. (2009), The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Springer Science & Business Media.
Hou, K., Xue, C. and Zhang, L. (2020), “Replicating Anomalies,” The Review of Financial Studies, 33(5), 2019–2133.
Huberman, G. and Kandel, S. (1987), “Mean-Variance Spanning,” The Journal of Finance,42(4), 873–888.
Jegadeesh, N. (1990), “Evidence of Predictable Behavior of Security Returns,” The Journal of Finance, 45(3), 881–898.
Jegadeesh, N. and Titman, S. (1993), “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency,” The Journal of Finance, 48(1), 65–91.
Lewellen, J. (2015), “The Cross-Section of Expected Stock Returns,” Critical Finance Review,4(1), 1–44.
Lintner, J. (1965), “The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets,” The Review of Economics and Statistics, 47(1),13–37.
Newey, W. K. and West, K. D. (1987), “A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix,” Econometrica, 55(3), 703–708.
Onatski, A. (2010), “Determining the Number of Factors from Empirical Distribution of Eigenvalues,”The Review of Economics and Statistics, 92(4), 1004–1016.
Sharpe, W. F. (1964), “Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk,” The Journal of Finance, 19(3), 425–442.
Sharpe, W. F. (1966), “Mutual Fund Performance,” The Journal of Business, 39(1), 119–138.
Uematsu, Y., Fan, Y., Chen, K., Lv, J. and Lin, W. (2019), “SOFAR: Large-Scale Association Network Learning,” IEEE Transactions on Information Theory, 65(8), 4924–4939.
Uematsu, Y. and Yamagata, T. (2023), “Estimation of Sparsity-Induced Weak Factor Models,”Journal of Business & Economic Statistics, 41(1), 213–227.
Wang, K.-Y., Jiang, C.-H. and Huang, Y.-S. (2009), “Market States and the Profitability of Momentum Strategies: Evidence from the Taiwan Stock Exchange,” The International Journal of Business and Finance Research, 3(1), 89–102.
Wei, J. and Zhang, Y. (2026), “Can Principal Component Analysis Preserve the Sparsity in Factor Loadings?” Econometric Theory, 1–24.
Zou, H. and Hastie, T. (2005), “Regularization and Variable Selection via the Elastic Net,”Journal of the Royal Statistical Society: Series B (Statistical Methodology), 67(2), 301–320.