Lightweight Ensemble Models for Emotion Analysis in E-Commerce User Reviews: Performance and Scalability ^*

This paper investigates the integration of sentiment analysis into emotion classification models applied to user reviews, aiming to enhance predictive performance and accuracy. While existing deep learning models, such as BERT, have achieved state-of-the-art results in various NLP tasks, their computational requirements can render them impractical for real-time applications. To address this challenge, a hybrid approach is proposed that combines the VADER sentiment analysis tool with the lightweight DistilBERT transformer model for classifying emotional sentiments in textual data. Utilizing a comprehensive dataset of labeled reviews, the study rigorously trains and evaluates the model’s capabilities. Experimental results reveal that enabling sentiment analysis significantly improves model performance, achieving an accuracy of 98.50%, precision of 91.50%, recall of 92.00%, and F1 score of 91.75%. These findings highlight the importance of sentiment scores in providing essential contextual information, which facilitates better handling of ambiguous language and enhances overall classification effectiveness.