Title: Sentiment analysis on stocks: a hybrid feature extraction technique on 14 classifiers

Authors: Meera George; R. Murugesan

Addresses: Department of Humanities and Social Sciences, National Institute of Technology, Tiruchirappalli – 620015, Tamil Nadu, India ' Department of Humanities and Social Sciences, National Institute of Technology, Tiruchirappalli – 620015, Tamil Nadu, India

Abstract: Accurately predicting stock prices is challenging and has garnered massive attention from researchers and investors alike. Though the literature has shown sentiment analysis as a promising approach for efficient stock price prediction, it has found a considerable gap in studies using multiple feature extraction techniques with hybrid models for the efficient sentiment classification. Under these circumstances, this study aims to perform sentiment analysis using five feature extraction techniques including a hybrid and 14 classifiers for the accurate classification of stock tweets. The study extracted 21,121 tweets spanning March 2022 to December 2022 using Twitter application programming interface. The empirical result shows the superiority of the hybrid feature extraction technique over the other methods. The support vector machine classifier with a hybrid feature extraction technique is found to be the best-performing sentiment analysis model for Twitter stock data. The study has potential applications in building optimal investment strategies and decision-making.

Keywords: stock price; sentiment analysis; classifiers; feature extraction; hybrid.

DOI: 10.1504/IJADS.2025.143083

International Journal of Applied Decision Sciences, 2025 Vol.18 No.1, pp.84 - 112

Received: 08 Jun 2023
Accepted: 08 Aug 2023
Published online: 03 Dec 2024 *

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