Title: Stock prediction using non-negative discriminative feature selection

Authors: Lawrence Oseremen Agbator; Benjawan Srisura

Addresses: Faculty of Science, Department of Computer Science, Edo State Polytechnic Usen, P.M.B. 1104 Benin City, Edo State, Nigeria ' Department of Information Technology, Vincent Mary School of Science and Technology, Assumption University of Thailand, 592/3 Soi Ramkhamhaeng 24, Ramkhamhaeng Rd., Hua Mak, Bang Kapi, Bangkok 10240, Thailand

Abstract: Feature subset selection in stock prediction is increasingly attracting more interest within the research community. This is because the movement of the stock price is influenced by many features in the stock market, and it is a challenging task to select the optimal predictive features. Feature selection is a pre-processing step of data mining that aims to justify predictive features from a given dataset. Using technical indicators (TIs) in stock market forecasting is widely adopted by investors and researchers. Using a minimal number of these technical indicators is required as a pivot task and used for almost all successful predictions. However, obtaining the predictive feature subset remains an area of active research. Using efficient feature selection will lead to the minimum number of relevant features and thus affect the prediction performance in diverse ways. Therefore, in this study, we intend to propose a feature selection that can lead to better prediction performance.

Keywords: fundamental information; technical information; feature selection; spectral analysis; Laplacian graph; NDFS; non-negative discriminative feature selection.

DOI: 10.1504/IJDATS.2022.128263

International Journal of Data Analysis Techniques and Strategies, 2022 Vol.14 No.3, pp.186 - 202

Received: 09 Jan 2022
Accepted: 04 Apr 2022
Published online: 16 Jan 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article