Sentiment Analysis Using Learning-based Approaches: A Comparative Study

Sentiment analysis, which involves analyzing text data and using language computation to extract valuable information, is a significant focus in Natural Language Processing (NLP). It is widely used in various applications such as product review analysis, customer feedback analysis, and social media monitoring. This research investigates the performance of different machine learning and deep learning models for sentiment analysis on a dataset of customer reviews from an e-commerce platform. A total of eight approaches have been presented in this study including LightGBM, SVM, KNN with bagging, MultinomialNB, DNN, LSTM, BERT, and RoBERTa. The performance for all the proposed models was compared using four evaluation metrics: accuracy, precision, recall and F1-score. The experimental results indicate that the SVM model has outperformed all the other methods with a testing accuracy of 73.98%. The F1-score, precision and recall are also the highest at 0.71, 0.72 and 0.70 respectively. This study contributes to the sentiment analysis literature by demonstrating the effectiveness of different models for sentiment analysis on customer reviews datasets.