2024 | Erkut Memiş, Hilal Akarkamçı (Kaya), Mustafa Yeniad, Javad Rahebi, Jose Manuel Lopez-Guede
This paper presents a comparative study of sentiment analysis for financial tweets using deep learning methods. The study focuses on Turkish financial tweets, which were collected and tagged as "positive," "negative," or "neutral." Binary and multi-class datasets were created, and word embedding and pre-trained word embedding techniques were used for tweet representation. Five deep learning models—Neural Network, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and GRU-CNN—were employed as classifiers. The results showed that the CNN model with pre-trained word embedding achieved the best performance, achieving 83.02% accuracy for binary classification and 72.73% for multi-class classification. The study highlights the importance of sentiment analysis in financial decision-making and provides insights into the practical applications of deep learning in natural language processing.This paper presents a comparative study of sentiment analysis for financial tweets using deep learning methods. The study focuses on Turkish financial tweets, which were collected and tagged as "positive," "negative," or "neutral." Binary and multi-class datasets were created, and word embedding and pre-trained word embedding techniques were used for tweet representation. Five deep learning models—Neural Network, Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), and GRU-CNN—were employed as classifiers. The results showed that the CNN model with pre-trained word embedding achieved the best performance, achieving 83.02% accuracy for binary classification and 72.73% for multi-class classification. The study highlights the importance of sentiment analysis in financial decision-making and provides insights into the practical applications of deep learning in natural language processing.