機器學習於投資組合報酬率之影響

Abstract

在投資風氣越來越濃厚的世代之中，投資逐漸成為每個人在生活之中，或多或少都會進行的事。在這樣的環境之下，如何選擇投資標的已成為人們關心的議題。而投資組合為進行投資時一個優良的選擇，可以依照投資者不同的投資目的來進行組合，而透過多個標的共同進行分散權重的投資，達到適當的風險分散。 本研究想找尋在投資組合的風險分散之下，如何穩定的建構出一個可以有高收益率的投資組合，藉此避免產生若僅投資單一標的，因標的特殊原因而導致巨大損益的情形。 我們透過基於回歸的模型，以及機器學習模型，依照各模型的特性進行適合其模型原理的投資組合建構，其中包含了常使用來進行投資組合建構的深度模型以及分群模型。而本研究與之前的不同之處在於，針對此兩種不同原理的模型，以及所創建出的不同含義之投資組合進行成效相比，同時衡量在不同的市場環境以及時間長度之下，成效是否有何變化。 最後，根據我們的實驗結果所顯示，深度模型中的Random Forest模型無論在何種的市場環境之下皆獲得了最優的實驗結果。且在長期的實驗之下，可獲得穩定之效益，因此由此實驗就可知，Random Forest模型在建構高收益率之投資組合上，有良好且穩定的效果。

In the generation with an increasingly investment ethos, investing has gradually become a part of everyone's life. In such an environment, how to choose investment targets has become a topic of concern. Portfolio is an excellent choice for investment. It can be combined according to the different investment purposes of investors, and through the joint investment of multiple targets to diversify the weights, to achieve appropriate risk diversification. This study seeks to find out how to construct a high-yield investment portfolio under the risk diversification of the portfolio, so as to avoid the situation that if only a single target is invested, huge losses will be caused due to the special reasons of the target. We use regression-based models and machine learning models to construct portfolios suitable for their model principles according to the characteristics of each model, including deep learning and clustering for portfolio construction. The difference between this study and the previous one is that the models based on these two different portfolios with different meanings created compare the performance, and at the same time measure whether the performance is under different market conditions and time periods. Finally, according to our experimental results, the Random Forest model has obtained the best experimental results in any market environment. And under long-term experiments, the models is still stable, so from this experiment, we can see that the Random Forest model has a good effect in constructing high-return investment portfolios.