Emotion-Based Movie Reviews Classification Using Deep Learning and Facial Image Recognition

The abstract highlights the significance of Facial Expression Recognition (FER) in deciphering human emotions, serving as a vital tool across various domains. Challenges like overfitting in modern deep FER systems are attributed to limited training data and extraneous factors. Integrating sentiment analysis with FER in face recognition video analysis aids in strategic decision-making for business development. The study proposes a dual-process system comprising face detection and FER, employing sentiment analysis to gauge responses to reviews. Results indicate that the proposed EfficientNet-B4 model achieves notable performance, attaining 91% and 98% accuracy on training and testing datasets respectively. EfficientNet-B4 methodology achieves notable accuracy in discerning emotions in movie reviews, surpassing existing models. The research contributes insights into advancing face recognition technologies for informed decision-making in business contexts.