<![CDATA[Various factors, including market perception reflected in media information, Influence crude oil price fluctuations. This study aims to analyse the Influence of news sentiment on crude oil price movements using a deep learning–based sentiment analysis approach. The dataset consists of 108 news headlines and daily closing oil prices from January to May 2025. It is important to note that this dataset is relatively small for deep learning models like LSTM, GRU, and BiLSTM, which constitutes a major constraint for this study. The news text was processed with case folding, tokenisation, stopword removal, and lemmatisation (not stemming to preserve semantic integrity for BERT), then automatically labelled using the DistilBERT model. The BERT-based vector representations were used as input for three classification models: LSTM, GRU, and BiLSTM. The evaluation results showed that all three models achieved the same average validation accuracy of 85.27%. However, the GRU model is identified as the optimal performer, achieving the lowest validation loss (0.3324), indicating better generalisation performance than LSTM and BiLSTM. Further analysis reveals that news sentiment tends to align with oil price trends, particularly during significant market shifts.]]>
Copyrights © 2026
