群眾智慧投資決策 – 配對交易、投資組合選擇以及股票價格/風險動向預測

Opinion-based Investment Decisions – Pair Trading, Portfolio Selection, and Stock Price/Risk Movement Prediction

許庭瑋(Ting-Wei Hsu)

指導教授 ： 陳信希

共同指導教授 ： 陳孟彰(Meng Chang Chen)

國立臺灣大學/電機資訊學院/資料科學學位學程/碩士(2021年)

https://doi.org/10.6342/NTU202102899

摘要

在進行投資決策時，投資者都會面臨著風險與收益的權衡。然而，以往在自然語言處理領域的工作大多集中在預測股票價格或波動率的走勢上，而沒有考慮其他投資議題。這篇論文介紹三個基於社群媒體意見的投資任務—配對交易、投資組合選擇以及股票價格/風險動向預測。首先，為了對沖市場風險，我們提出了一種基於社群媒體的配對交易策略。與先前的任務設置相比較，我們的實驗結果表明，採用配對任務設置的神經網絡模型在準確性和盈利性指標上均都有較好的表現。第二，很少有研究在處理投資組合選擇時考慮金融界的非結構化數據。我們引入了一種新穎的基於財務文本的投資組合選擇任務，並提出了新的目標函數去處理投資者不同的風險偏好。同時討論了夏普比率和波動率兩個指標對選擇投資組合的實證研究。第三，我們提出了語義保留的增廣方法，並在六個公開資料集上均達到更好的表現，且據此來更精準地預測金融市場未來股票的價格與風險動向。此外，我們將以上研究成果發展成展示網站，提供投資者財務決策上的建議。綜上所述，本研究為未來基於財金社群媒體的群眾智慧投資決策引入了新的研究方向。

關鍵字

意見探勘 ； 自然語言處理 ； 財金社群媒體

並列摘要

Every investor faces a risk­return tradeoff when making investment decisions. However, most of the previous works in Natural Language Processing focus on predicting the movement of stocks’ prices or volatilities but do not consider the other investment issues. This thesis presents three investment tasks based on social media opinions. Firstly, we propose an idea of opinion­based pair trading. Comparing with the task settings of the previous works, our experimental results show that the pairwise task setting performs better in both accuracy and profitability metrics. Secondly, few works consider unstructured data in the financial community when dealing with portfolio selection. We introduce a novel opinion­based portfolio selection task and propose new objective functions presenting different risk appetites of investors. Thirdly, we propose a semantics­preserved augmentation approach without altering the aspect­level polarity. Our approach achieves better performances on six publicly available sentiment analysis datasets and the real-world stock price/risk movement prediction scenarios. In addition, we have developed a demonstration website to give advice to investors on financial decisions. In summary, this thesis introduces new directions for future investment research based on the opinions on social media platforms.

並列關鍵字

Opinion Mining ； Natural Language Processing ； Financial Social Media

參考文獻

[1] Ramit Sawhney, Shivam Agarwal, Arnav Wadhwa, and Rajiv Ratn Shah. Deep attentive learning for stock movement prediction from social media text and company correlations. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP).

[2] Yumo Xu and Shay B. Cohen. Stock movement prediction from tweets and historical prices. In ACL, July 2018.

[3] Qikai Liu, Xiang Cheng, Sen Su, and Shuguang Zhu. Hierarchical complementary attention network for predicting stock price movements with news. New York, NY, USA, 2018. Association for Computing Machinery.

[4] Ziniu Hu, Weiqing Liu, Jiang Bian, Xuanzhe Liu, and Tie­Yan Liu. Listening to Chaotic Whispers: A Deep Learning Framework for News­oriented Stock Trend Prediction. In WSDM, 2018.

[5] Yu Qin and Yi Yang. What you say and how you say it matters: Predicting stock volatility using verbal and vocal cues. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 390–401, Florence, Italy, July 2019. Association for Computational Linguistics.