基於密集隱性相關新聞序列之股票趨勢預測

Abstract

大量網路新聞分析有利於股市趨勢的預測，先前的研究往往只考慮有提及公司或股票名稱的新聞作為分析的對象，但是並非每間公司每天都會被新聞報導。如果只考慮具有上述特徵的新聞，容易造成資料的稀疏。同時通過產業相關新聞的分析，做為財務決定之參考較符合真實人類的行為。因此，本研究將使用無明確提及公司名稱之新聞，即隱性相關新聞，來豐富預測模型訓練之資料來源。為能充分利用隱性相關新聞，本研究提出基於新聞篩選機制的深度神經網絡框架，該框架結合文件表徵學習和偕同過濾來捕捉新聞和股票之間的關係。本研究實驗於真實世界資料，並且與最先進的模型比較，實驗結果顯示有效性以及準確率都有提升。

Analyzing online news content benefits stock price trend prediction. Previous studies on news-oriented stock market prediction focus mainly on news with explicit stock mentions for a specific prediction target, and may suffer from data sparsity. As taking into consideration other related news - e.g., sector-related news - is a crucial part of real-world decision-making, we explore the use of news without explicit target mentions to enrich the information for the prediction model. We first conduct an empirical analysis on real-world news collected from a well-known financial website. To leverage implicit-related information, we devise a neural network framework that jointly learns with a news selection mechanism to extract implicit news from the chaotic daily news pool. The news distilling network (NDN), our proposed model, takes advantage of neural representation learning and collaborative filtering to capture the relationship between stocks and news. With NDN, we learn latent stock and news representations to facilitate similarity measurements, and apply a gating mechanism to prevent noisy news representations from flowing to a higher level encoding stage, which encodes each day's selected news representation. Extensive experiments on real-world stock market data demonstrate the effectiveness of our framework and show improvements over state-of-the-art techniques.