Sentiment analysis of financial news using the bert model

Abstract

Financial decisions are strongly reliant on correct sentiment analysis. Traditional methods frequently fall short of capturing hidden sentiments. A financial market dataset comprising 5,842 reviews was collected for analysis. Among these, 1,852 reviews were positive, 860 were negative, and 3,130 were neutral. After downsampling, the data was divided into two groups: the training set and the test set. The training set was employed to train the model, while the test set was reserved for evaluation. This study applies a deep learning approach using the Bidirectional Encoder Representations from Transformers (BERT) model to train the dataset. The performance of the model was measured using accuracy, precision, recall, and F1-score. The model gave a high performance with an accuracy of 95.29%, precision of 95.37%, recall of 95.24%, and a minimal loss of 9.07%. Notably, the F1-score, which provides a balanced evaluation of the model's efficiency, is 95.32%. These findings highlight the BERT model's effectiveness in conducting sentiment analysis in financial markets. The study not only enhances the field of financial sentiment analysis, but it also emphasises the practicality and dependability of using deep learning techniques to extract significant insights from financial data.