Volume 23, Issue 60, July - December, 2026

A Polymorphic Graph Attention Bayesian Asymmetric Quantized Neural Network-Based Approach for Sentiment Analysis of Movie Reviews

Vinoth R^1♦, Sudha K², Lakshmipriya C³, Udhayashankar S⁴

¹Dept of CSE, RMK College of Engineering and Technology, India
²Dept of CSBS, RMD Engineering College, India
³Dept of CSE, SA Engineering College, India
⁴Dept of AI&DS, Vel Tech Multi Tech Dr. Rangarajan Dr. Sakunthala Engineering College, India

^♦Corresponding Author
Vinoth R, Dept of CSE, RMK College of Engineering and Technology, India

ABSTRACT

Analyzing user sentiment from movie reviews requires deep learning techniques because the large volume of reviews makes manual analysis is impractical. Sentiment analysis is now essential for understanding public opinion because people use online platforms to share vast amounts of data and opinions. Traditional approaches are capable of dealing with limited information, while big and complex data need sophisticated techniques. In this study, a novel approach is introduced for movie review classification based on sentiments using Polymorphic Graph Attention Bayesian Asymmetric Quantized Neural Network with Emperor Penguin Optimization (PGABAQNNet-EPO). Performance analysis of the proposed model is done based on three commonly used databases, which are MovieLens, IMDb, and Rotten Tomatoes. The results show a high accuracy rate of 99.1%, 99.4%, and 99.3% for MovieLens, IMDb, and Rotten Tomatoes, respectively, and an F1-score of more than 99%, the results demonstrate outstanding performance. The model achieves rapid computation time of 0.02 second along with low error values. The algorithm shows strong capabilities for sentiment analysis and provides a reliable approach to classify movie reviews with high speed.

Keywords: Bayesian asymmetric quantized neural networks, Dual-domain Cswin transformer, Emperor Penguin Optimization, natural language processing, Polymorphic graph attention.

Indian Journal of Engineering, 2026, 23(60), e12ije1729

PDF

DOI: https://doi.org/10.54905/disssi.v23i60.e12ije1729

Published: 15 September 2026