本研究旨在提出一種整合多特徵時間序列相似性之股票分群方法，結合機器學習建立群內訓練股價預測模型，並且進一步將預測結果應用於投資組合建構。本文使用臺灣上市公司自2019年至2024年之股價與財務資料，先透過動態時間校正距離與歐式距離衡量多個價量資訊或財務比率的時間序列相似性，並以k-medoids演算法進行股票分群，搭配滾動式更新機制反映市場結構之變化，接著於各群內訓練長短期記憶網路（LSTM）模型預測隔日與20日後收盤價，並根據預測結果挑選預期報酬最高之個股依三種不同資產配置策略建構投資組合，回測其績效。
實證結果顯示，分群後群內訓練模型在預測準確性方面整體優於不分群模型，能有效降低預測誤差，尤其在以價量資訊分群並設定分5群時表現最佳，回測績效方面，分群模型所建構之投資組合於多數情境皆能優於基準模型與大盤。綜合而言，本研究驗證多特徵時間序列分群方法能有效降低資料異質性，提升預測穩定性與投資表現，為股價預測與選股策略提供可行的整合方法。

This study proposes a stock clustering framework based on the similarity of multivariate time series features, which is integrated with deep learning models for within-cluster stock price prediction. The predictions are further applied to stock selection and portfolio construction. Using data from Taiwan-listed companies between 2019 to 2024, the study computes time series similarity—based on either daily price-volume data or quarterly financial ratios—using Dynamic Time Warping and Euclidean Distance, and applies the k-medoids algorithm for clustering. A rolling mechanism updates cluster assignments annually to reflect structural market changes. Within each cluster, Long Short-Term Memory models are trained to forecast stock prices for the next day and 20 days ahead. Predicted returns are ranked to identify top-performing stocks, which are then used to construct investment portfolios under three asset allocation strategies.
Empirical results show that clustered models outperform non-clustered baselines in forecasting accuracy, especially when using price-volume features with five clusters. In terms of investment performance, portfolios constructed from the clustered models consistently outperform benchmark models and the market index across most scenarios. Overall, the study demonstrates that multivariate time series clustering effectively reduces data heterogeneity, enhances predictive stability, and improves practical investment outcomes.

Aithal, P. K., Geetha, M., U, D., Savitha, B., & Menon, P. (2023). Real-Time Portfolio Management System Utilizing Machine Learning Techniques. IEEE Access, 11, 32595–32608. https://doi.org/10.1109/access.2023.3263260
Amini, S., Hudson, R., Urquhart, A., & Wang, J. (2021). Nonlinearity everywhere: implications for empirical finance, technical analysis and value at risk. The European Journal of Finance, 27(13), 1326–1349. https://doi.org/10.1080/1351847x.2021.1900888
Atsalakis, G. S., & Valavanis, K. P. (2009). Surveying stock market forecasting techniques – Part II: Soft computing methods. Expert Systems with Applications, 36(3), 5932–5941. https://doi.org/10.1016/j.eswa.2008.07.006
Bandara, K., Bergmeir, C., & Smyl, S. (2020). Forecasting across time series databases using recurrent neural networks on groups of similar series: A clustering approach. Expert Systems with Applications, 140. https://doi.org/10.1016/j.eswa.2019.112896
Chaweewanchon, A., & Chaysiri, R. (2022). Markowitz Mean-Variance Portfolio Optimization with Predictive Stock Selection Using Machine Learning. International Journal of Financial Studies, 10(3). https://doi.org/10.3390/ijfs10030064
Hyndman, R. J., Wang, E., & Laptev, N. (2015). Large-Scale Unusual Time Series Detection 2015 IEEE International Conference on Data Mining Workshop (ICDMW),
Li, M., Zhu, Y., Shen, Y., & Angelova, M. (2023). Clustering-enhanced stock price prediction using deep learning. World Wide Web, 26(1), 207–232. https://doi.org/10.1007/s11280-021-01003-0
Markowitz, H. (1952). Portfolio Selection. The Journal of Finance, 7(1), 77–91. https://doi.org/10.2307/2975974
Medarhri, I., Hosni, M., Nouisser, N., Chakroun, F., & Najib, K. (2022). Predicting Stock Market Price Movement using Machine Learning Techniques 2022 8th International Conference on Optimization and Applications (ICOA),
Phuoc, T., Anh, P. T. K., Tam, P. H., & Nguyen, C. V. (2024). Applying machine learning algorithms to predict the stock price trend in the stock market – The case of Vietnam. Humanities and Social Sciences Communications, 11(1). https://doi.org/10.1057/s41599-024-02807-x
Savitzky, A. (1964). Smoothing and differentiation of data by simplified least squares procedures. Analytical chemistry, 36(8), 1627–1639. https://doi.org/10.1021/ac60214a047
Shynkevich, Y., McGinnity, T. M., Coleman, S. A., Belatreche, A., & Li, Y. (2017). Forecasting price movements using technical indicators: Investigating the impact of varying input window length. Neurocomputing, 264, 71–88. https://doi.org/10.1016/j.neucom.2016.11.095
Vásquez Sáenz, J., Quiroga, F. M., & Bariviera, A. F. (2023). Data vs. information: Using clustering techniques to enhance stock returns forecasting. International Review of Financial Analysis, 88. https://doi.org/10.1016/j.irfa.2023.102657
Wang, X., Yang, K., & Liu, T. (2021). Stock Price Prediction Based on Morphological Similarity Clustering and Hierarchical Temporal Memory. IEEE Access, 9, 67241–67248. https://doi.org/10.1109/access.2021.3077004
Warren Liao, T. (2005). Clustering of time series data—a survey. Pattern Recognition, 38(11), 1857–1874. https://doi.org/10.1016/j.patcog.2005.01.025
Yu, P., & Yan, X. (2019). Stock price prediction based on deep neural networks. Neural Computing and Applications, 32(6), 1609–1628. https://doi.org/10.1007/s00521-019-04212-x