Active Learning Empowered Sentiment Analysis System for E-Commerce Product Reviews

The paper follows the Active Learning framework with a view to improving model performance on a small amount of labeled data.Feature extraction in text shall be through TF-IDF, coupled with logistic regression as the main classifier.User reviews through the web will be treated at the back end, classifying the textual data followed by visual analytics in the form of pie charts, bar graphs, and line plots to interpret, effectively the sentiment trends by customers and administrators.Active learning is an approach toward improving classifier accuracy through effective sampling of uncertain samples for the retraining of the classifier iteratively.The proposed system is a major enhancement to review transparency, product evaluation, and decisions related to the e-commerce environment.Results indicate improvement in accuracy and interpretability concerning nonactive learning methods.