Open AccessNews-based Machine Learning and Deep Learning Methods for Stock Prediction
Author(s) -
Junjie Guo,
Bradford Tuckfield
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1642/1/012014
Subject(s) - artificial intelligence , stock (firearms) , machine learning , deep learning , computer science , stock market , stock market prediction , econometrics , economics , engineering , geography , mechanical engineering , context (archaeology) , archaeology
Stocks occupy a vital position in the financial market. Over the years, scholars have made unremitting efforts in forecasting the stock market. Because the more accurate the prediction, the more people will profit from the stock market. Machine learning has achieved excellent results in stock prediction. Nowadays, with the rise of deep learning, the stock prediction methods used by people are beginning to lean towards deep learning, and many results have been achieved. This paper will use news rather than traditional stock structured data for stock prediction, and we will use machine learning and deep learning methods in contrast. Moreover, we use natural language processing to process the news. The objects of prediction are stock indexes (DJIA, S & P500) and individual stocks (IBM, JPM). We find that deep learning performs at least 4.5% better than machine learning on prediction tasks related to stock indexes, and at least 3% better in the prediction of individual stocks. We discuss the implications of this result.