在過去投資組合管理的研究上，多是利用傳統的投資組合理論進行資產管理，而近年來越來越多的研究將人工智能用於優化投資組合，本文嘗試以元大台灣50 ETF之成分股為基礎，結合長時間短期記憶神經網路來預測股價的下期變動，再從中選取預測報酬率較高的多檔股票完成股票篩選，最後結合三種資產權重配置的方法建構風險分散投資組合，分別為等權重、選取最小變異數的平均數-變異數模型以及考慮了預測報酬率的MVF模型。
實證結果發現在大部分的時間段裡，運用長時間短期記憶神經網路進行選股的交易策略皆呈現較高的報酬，無論是結合等權重、平均數-變異數模型及MVF的方法進行資產權重配置，投資績效皆能優於元大台灣50 ETF，顯示長時間短期記憶神經網路在投資組合當中具有一定程度的貢獻，而在實際交易情況下，則是結合等權重、平均數-變異數模型呈現較好之績效。

In the past research on investment portfolio management, most of them used traditional investment portfolio theory for asset management. In recent years, more and more researches have used artificial intelligence to optimize investment portfolios. In this thesis, we try to use the constituent stocks of the Yuanta/P-shares Taiwan Top 50 ETF as the basis and combine the Long-Short Term Model to predict the next stock price change. Then select multiple stocks with higher predicted returns. Finally, three asset weight allocation methods are combined to construct a risk-diversified investment portfolio, which are equal weight, the mean-variance model with the minimum variance selected, and the Mean–variance with forecasting model considering the predicted rate of return.
The empirical results found that the trading strategy of using Long-Short Term Model for stock selection has higher returns in most time periods. Regardless of the method of asset weight allocation combined with equal weight, mean-variance model and Mean–variance with forecasting model, the investment performance can be better than Yuanta/P-shares Taiwan Top 50 ETF. Shows that the Long-Short Term Model has a certain degree of contribution in the investment portfolio. In the actual trading situation, the combination of equal weight and mean-variance model presents better performance.

[1] Chen, Y. J., & Hao, Y. J. (2018). Integrating principle component analysis and weighted support vector machine for stock trading signals prediction. Neuro computing, (321), 381–402.
[2] Chong, E., Han, C., & Park, F. C. (2017). Deep learning networks for stock market analysis and prediction: Methodology, data representations, and case studies. Expert Systems With Applications, (83), 187–205.
[3] Ding, X., Zhang, Y., Liu, T., & Duan, J. (2015). Deep learning for event-driven stock prediction. Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence.
[4] Fischer, T., & Krauss, C. (2018). Deep learning with long short-term memory networks for financial market predictions. European Journal of Operational Research, 270, 654–669
[5] Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735–1780.
[6] Kara, Y., Boyacioglu, M. A., & Baykan, O. K. (2011). Predicting direction of stock price index movement using artificial neural networks and support vector machines: The sample of the istanbul stock exchange. Expert Systems with Applications, (38), 5311–5319.
[7] Kraus, M., & Feuerriegel, S. (2017). Decision support from financial disclosures with deep neural networks and transfer learning. Decision Support Systems, (104), 38–48.
[8] Ma, Y., Han, R., & Wang, W. (2021). Portfolio optimization with return prediction using deep learning and machine learning. Expert Systems With Applications 165 113973.
[9] Markowitz, H. (1952). Portfolio selection. The Journal of Finance, 7(1), 77-91.
[10] Paiva, F. D. A, Cardoso, R. T. N., Hanaoka, G. P., & Duarte, W. M. (2019). Decision– making for financial trading: A fusion approach of machine learning and portfolio selection. Expert Systems with Applications, (115), 635–655.
[11] Patel, J., Shah, S., & Thakkar, P. (2015). Predicting stock and stock price index movement using Trend Deterministic Data Preparation and machine learning techniques. Expert Systems with Applications, 42, 259–268.
[12] Rather, A. M., Agarwal, A., & Sastry, V. N. (2015). Recurrent neural network and a hybrid model for prediction of stock returns. Expert Systems with Applications, 42, 3234–3241.
[13] Wang, W., Li, W., Zhang, N., & Liu, K. (2020). Portfolio formation with preselection using deep learning from long-term financial data. Expert Systems With Applications 143 113042.