建構輔以機器學習的技術交易動能策略

金融時間序列的非定態特性使得預測未來股價非常困難。但精準預測股價並不是投資獲利的唯一方法。股價趨勢相對容易掌握，即交易訊號的預測較易被實踐。只要投資人可以精準擇時，在正確的時間點進行買賣交易，皆可因此而獲利。本研究以邏輯斯回歸演算法加入技術交易與投資人情緒指標建構機器學習模型產生交易訊號，希望藉由不同特徵值提升模型之市場擇時能力，以正確捕捉股市動能。另外，本研究亦針對空頭市場時各策略之表現以及策略是否有降低投資風險的效果進行討論，並針對捕捉不同天期之動能探討策略績效。

關鍵字

機器學習；邏輯斯回歸演算法；技術交易；投資人情緒指標；動能策略；交易訊號；市場擇時

並列摘要

The non-stationary feature of financial time series makes the prediction of future stock price harder. However, predicting stock price correctly is not the only way to get return from investing. The trend of stock price is easier to control, which means predicting trading signals is likely to put to practice. As long as investor is able to time the market perfectly and make the right trading decision, making profit is no longer difficult. In our study, we apply technical trading indicators and investors sentiment indicator to logistic regression algorithm to build a machine learning model in order to predict trading signals. We intend to improve the model ability of timing market via importing different features. Furthermore, we discuss about the performance of different strategies and if they lower down the investment risk when facing bear market. We also talk about the performance of different strategies by capturing momentum of different time horizons in further discussion.

並列關鍵字

Machine Learning ； Logistic regression algorithm ； Technical trading ； Investor sentiment indicator ； Momentum strategy ； Trading signal ； Market timing

參考文獻

1. Barberis, Nicholas, Andrei Shleifer, and Robert Vishny. 1998, A Model of Investor Sentiment, Journal of Financial Economics 49 (3): 307-343

Google Scholar

2. Daniel, K. D., D. Hirshleifer, and A. Subrahmanyam, 1998, Investor psychology and security market under- and over-reactions, Journal of Finance 53, 1839–1886.