| Sentiment analysis has become increasingly important for extracting insights from user opinions and emotions, particularly in applications like online product reviews and social media analytics. However, traditional approaches often struggle with difficulties like the complexity of natural language, context, and imbalanced datasets. To overcome these issues, we provide a framework that is hybrid and incorporates Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, and eXtreme Gradient Boosting (XGBoost) to improve the accuracy and robustness of sentiment classification tasks. CNNs extract key local features from text, which are then processed by LSTMs to capture sequential dependencies and contextual nuances. To enhance prediction accuracy, XGBoost refines its classifications by recognising complex patterns in the data. We tested our proposed model using the IMDB movie review dataset, where it consistently outperformed individual models across various performance metrics, including accuracy, precision, recall, F1-score, and AUC-ROC. These findings demonstrate the advantages of integrating deep learning with gradient boosting for sentiment analysis, providing a more reliable and adaptable approach compared to traditional methods. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.