Improved Ensemble Learning Technique for Classification of Movie Reviews

View at publisher Summarise with VIRAT
Authors: Akash Mishra, Hrishikesh Nirale, Pratham Patel, Nilesh Yadav
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Sentiment analysis, a fundamental application of Natural Language Processing (NLP), employs machine learning to recognize and categorize opinions within text. As user-generated data explodes, accurate sentiment detection is essential in industries like finance, e-commerce, and social media. Ensemble methods become increasingly popular with their ability to tap the strengths of models and improve predictive power. This paper proposes a hybrid ensemble of BERT, LSTM, GRU, and a Flatten-based neural network, designed to capture contextual, sequential, and semantic features to enhance sentiment classification accuracy. Conventional sentiment analysis models are likely to fail when dealing with contextual subtlety, irony, and domain portability. Single-model architectures are proven to have limitations in scalability and real-time adaptability, which decrease their performance on dynamic or large datasets. These weaknesses emphasize the need for more robust, ensemble-based approaches that are accurate. The first, EDSA-Ensemble, presents a multi-classifier method specifically designed for event-based sentiment analysis. The meta-classifiers used were multiple ML models which gave variable accuracies, but the one with the highest accuracy was to be considered and which was Random Forest.

Fields of Study & Subjects

Sentiment Analysis and Opinion Mining Topic Modeling Stock Market Forecasting Methods

Publication details

Journal / source
CRC Press eBooks
Pages
288–291
Publication date
25 March 2026
Language
EN
Document type
Book chapter