TY - CHAP
T1 - Thai Stock Price Prediction from Daily News Contents
AU - Wattanakul, Pavit
AU - Phienthrakul, Tanasanee
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2022
Y1 - 2022
N2 - Stock market is a major place for investment. Many investors would like to accurately predict the stock price in order to get more profit. Stock price prediction by machine learning techniques is widely used according to an uncertainty movement of the stock price. However, this should be better if the related information is combined. Daily news is a source of information that reflects many factors. Both good and bad factors are indicated in the news. In this research, the news is used to forecast the stock price. Text processing is applied to extract word features. Then, word features are analyzed and selected. The selected features are used in prediction process. Three predictive models, random forest (RF), AdaBoost and support vector regression (SVR), are compared. SVR yields the best result with the lowest root mean square error (RMSE). The result shows that on 50 features from Spearman ranking, SVR is able to reduce the root mean-squared error by 8.63%.
AB - Stock market is a major place for investment. Many investors would like to accurately predict the stock price in order to get more profit. Stock price prediction by machine learning techniques is widely used according to an uncertainty movement of the stock price. However, this should be better if the related information is combined. Daily news is a source of information that reflects many factors. Both good and bad factors are indicated in the news. In this research, the news is used to forecast the stock price. Text processing is applied to extract word features. Then, word features are analyzed and selected. The selected features are used in prediction process. Three predictive models, random forest (RF), AdaBoost and support vector regression (SVR), are compared. SVR yields the best result with the lowest root mean square error (RMSE). The result shows that on 50 features from Spearman ranking, SVR is able to reduce the root mean-squared error by 8.63%.
KW - AdaBoost
KW - Daily news
KW - Feature ranking
KW - Random forest
KW - Stock price prediction
KW - Support vector regression
UR - https://www.scopus.com/pages/publications/85112018305
U2 - 10.1007/978-981-16-0739-4_80
DO - 10.1007/978-981-16-0739-4_80
M3 - Chapter
AN - SCOPUS:85112018305
T3 - Lecture Notes in Networks and Systems
SP - 851
EP - 861
BT - Lecture Notes in Networks and Systems
PB - Springer Science and Business Media Deutschland GmbH
ER -