TY - JOUR
T1 - Use GBDT to Predict the Stock Market
AU - Yang, Jin Shan
AU - Zhao, Chen Yue
AU - Yu, Hao Tong
AU - Chen, He Yang
N1 - Publisher Copyright:
© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
PY - 2020
Y1 - 2020
N2 - Prediction of stock market is important and popular for investors, to decrease the loss and increase the profit, how to prevent the risk becomes a special field and there is amount of solutions of the problem. Previous researches were focus on the factors that may influence the emotions of investors, the researchers finished researches based on the social media, the period of stock market and applied different models to extract the feature of stock. We first used natural language processing to deal with the text, including TF-IDF, Word2Vec, CountVectorizer and Doc2Vec. Then, we reduce dimension by PCA, finally, we used machine learning to deal with data, including Adaboost, XGboost and GBDT, decision tree, logistics regression. After voting process, we find GBDT is the most accurate algorithm to predict the stock. Comparing with previous work, we focus on the emotion and select information from news, which is more accurate than other researches.
AB - Prediction of stock market is important and popular for investors, to decrease the loss and increase the profit, how to prevent the risk becomes a special field and there is amount of solutions of the problem. Previous researches were focus on the factors that may influence the emotions of investors, the researchers finished researches based on the social media, the period of stock market and applied different models to extract the feature of stock. We first used natural language processing to deal with the text, including TF-IDF, Word2Vec, CountVectorizer and Doc2Vec. Then, we reduce dimension by PCA, finally, we used machine learning to deal with data, including Adaboost, XGboost and GBDT, decision tree, logistics regression. After voting process, we find GBDT is the most accurate algorithm to predict the stock. Comparing with previous work, we focus on the emotion and select information from news, which is more accurate than other researches.
KW - GBDT
KW - Natural language processing
KW - Stock market
KW - XGboost
UR - http://www.scopus.com/inward/record.url?scp=85100405860&partnerID=8YFLogxK
U2 - 10.1016/j.procs.2020.06.071
DO - 10.1016/j.procs.2020.06.071
M3 - Conference article
AN - SCOPUS:85100405860
SN - 1877-0509
VL - 174
SP - 161
EP - 171
JO - Procedia Computer Science
JF - Procedia Computer Science
T2 - 8th International Conference on Identification, Information and Knowledge in the Internet of Things, IIKI 2019
Y2 - 25 October 2019 through 27 October 2019
ER -